5% CO₂ is a potent, fast acting inhalation anticonvulsant

Abstract

Purpose

CO₂ has been long recognized for its anticonvulsant properties. We aimed to determine whether inhaling 5% CO₂ can be used to suppress seizures in epilepsy patients. The effect of CO₂ on cortical epileptic activity accompanying behavioral seizures was studied in rats and a non-human primate and based on these data, preliminary tests were carried out in humans.

Methods

In freely moving rats, cortical afterdischarges paralleled by myoclonic convulsions were evoked by sensorimotor cortex stimulation. 5% CO₂ was applied for 5 minutes, 3 minutes before stimulation. In macaque monkeys, hypercarbia was induced by hypoventilation while seizure activity was electrically or chemically evoked in the sensorimotor cortex. Seven patients with drug-resistant partial epilepsy were examined with video-EEG and received 5% CO₂ in medical carbogen shortly after electrographic seizure onset.

Results

In rats, 5% CO₂ strongly suppressed cortical afterdischarges, by ca. 75%, while responses to single-pulse stimulation were reduced by about 15% only. In macaques, increasing pCO₂ from 37 to 44-45 mmHg (corresponding to inhalation of 5% CO₂ or less) suppressed stimulation-induced cortical afterdischarges by about 70% and single, bicuculline-induced epileptiform spikes by ca. 25%. In a pilot trial carried out in 7 patients, a rapid termination of electrographic seizures was seen despite the fact that the application of 5% CO₂ was started after seizure generalization.

Conclusions

5% CO₂ has a fast and potent anticonvulsant action. The present data suggest that medical carbogen with 5% CO₂ can be used for acute treatment to suppress seizures in epilepsy patients.

Introduction

The anticonvulsant action of CO₂ was demonstrated as early as 1928 in humans with petit mal epilepsy by the suppression of the characteristic electroencephalogram (EEG) spike-wave pattern during administration of carbogen containing 10% CO₂ (Lennox, 1928; Lennox et al., 1936). This finding was extended in several studies on rodent, canine and non-human primate models (Pollock, 1949; Stein and Pollock, 1949; Dahlberg-Parrow, 1951; Woodbury et al., 1955; Mitchell and Grubbs, 1956; Woodbury et al., 1958; Meyer et al., 1961; Caspers and Speckmann, 1972; Balestrino and Somjen, 1988; Ziemann et al., 2008), and by observations in psychiatric patients where 15-30% CO₂ prevented electrically-induced convulsions (Pollock et al., 1949). In most of these studies relatively high CO₂ levels (around 10% and above) were used. Surprisingly, it has never been investigated whether standard medical carbogen with a low level of CO₂ (5%) could be used to suppress seizures in patients with epilepsy.

The present study was designed to determine the anticonvulsant properties of 5% CO₂ in adult animal seizure models and in humans with partial epilepsy. We studied cortical epileptic activity in unanesthetized, freely behaving rats (Kubova et al., 1996), anesthetized macaque monkeys (Haglund and Hochman, 2007) and in patients. In both animal models, 5% CO₂ (or the near-equivalent level reached by controlled hypoventilation in macaques) strongly suppressed cortical afterdischarges and associated behavioral seizures. Suppression of cortical activity by CO₂ was seen rapidly, within 1-2 minutes. Finally, we tested the effect of medical carbogen (5% CO₂ + 95% O₂) on the duration of spontaneous seizures in a pilot trial in 7 humans with drug-resistant epilepsy and found that inhalation of 5% CO₂ significantly reduces seizure duration even when applied after generalization.

Methods

Rat model of myoclonic epilepsy

Experiments were performed on 26 male Wistar rats (weight 270-350 grams) and were approved by the Animal Care Committee of the Institute of Physiology in agreement with the Czech Republic Animal Protection Law and European Community Council Directives 86/609/EEC. A rat model of myoclonic seizures was used in which cortical stimulation elicits bilateral epileptiform afterdischarges (ADs) in the EEG paralleled by clonic convulsions (Kubova et al., 1996). Silver electrodes were implanted epidurally under isoflurane anesthesia. A bipolar stimulation electrode was placed over the right somatosensory cortex (AP= +1 anterior and -1 mm posterior in relation to bregma, ML = 2.5 mm). Recording electrodes were positioned over the left somatosensory cortex (AP = 0 mm; ML = 2.5 mm), left parietal cortex (AP = -3 mm; ML = 3 mm), left occipital and right occipital cortex (AP= -6 mm; ML = 4 mm). A reference and a ground electrode were placed over the cerebellum. Electrodes were fixed to the skull with dental acrylic. Animals recovered for 1 week after surgery. Separate groups of animals were used for tests with 5% CO₂ (n=8 rats), 10% CO₂ (n=8) and acetazolamide (n=10). Experiments were performed at 25 °C. Rats were placed in an acrylic box (18×28×35 cm; supplied with a gas inlet on the bottom and outlet on the top of the box) 10 min before recordings and connected to a custom-made cable for EEG-recordings which was led through the gas outlet hole (Ø 1.5 cm) in the top lid of the box. EEG data were acquired at 2 kHz and filtered at 2-500 Hz (RA16PA preamplifier and Pentusa Base Station, Tucker-Davis Technologies, FL, USA). Single-pulse and AD stimulations were performed in freely moving rats (Kubova et al., 1996; Tsenov and Mares, 2007; Tsenov et al., 2009).

Single responses were evoked with 1 ms biphasic pulses using a custom-designed constant-current stimulator and were recorded 1-3 days before the AD experiments. Single-pulse thresholds ranged from 0.5 mA to 2 mA. To assess the effect of CO₂, EEG-recordings were made for 15 minutes with pulses given every 10 seconds at 2× threshold intensity. 5% CO₂ (in artificial air, with constant 18% O₂ in N₂; Linde Gas, Czech Republic) was applied for 5 minutes from the 5^th to the 10^th min at a constant high flow rate of about 20-25 liters/min, preventing possible leakage of air into the box. Capnographic measurements (Aseko Air XX infrared gas analyzer) verified that the CO₂ in the chamber reached the desired constant level within 15 seconds. The flow of carbogen did not disturb the rats behaviorally, apart from inducing some sniffing and exploratory behavior that could typically be observed during the first 30 seconds of exposure.

ADs were elicited by a 15 sec train of 1 ms biphasic pulses at 8 Hz, with threshold intensities from 1.5 mA to 7.5 mA. The intensity of behavioral seizures was quantified using a modified Racine's scale (Kubova et al., 1996). EEG was recorded continuously and behavior was videotaped. Each session consisted of 4 AD stimulations at 20 min intervals. The first stimulation was a test stimulation that is not included in the Results. Gas mixtures consisted of 5% or 10% CO₂ in artificial air (see above) and were applied 3 min before the second AD stimulation for a period of 5 min. Control sessions consisted of four AD stimulations in normal air. Acetazolamide (50 mg/kg; Sigma-Aldrich; freshly dissolved at 20 mM in 0.9% NaCl during careful titration to pH 8.5) or vehicle was administered i.p. 5 minutes after the first AD. Five AD stimulations were performed at 20 min intervals to assess the time-course of the drug effect. For both CO₂ and acetazolamide tests, half of the animals received the control treatment first. After an interval of at least 48 hours, control and experimental groups were reversed.

Data were analyzed offline using Matlab software (Mathworks, Inc., Natick, MA). For analysis of single evoked responses, the amplitude from the first negative wave (N1) to the following positive wave (P2) was measured (Fig 2; (Tsenov and Mares, 2007)). Values were averaged for 30 seconds bins (i.e. 3 responses). For each animal, N1P2 values were calculated as a percentage of the mean N1P2 value during the 5 minutes under control conditions. To compare responses during control conditions in air with those during 5% CO₂ application, the mean of the 30 sec bin values from the different animals was calculated, which was averaged for 10 consecutive 30 sec-bin values for control conditions, during application of CO₂ (taken from 30 sec after start of the CO₂ application), and for 8 consecutive 30 sec-bin values during recovery (starting from the 2^nd minute upon return to normocapnia).

Fig. 2.

5% CO₂ rapidly reduces somatosensory cortical responses to single pulse stimulation in freely behaving rats. Insets show example responses during the control condition, 5% CO₂ and recovery; leftmost trace depicts the measurement of the N1P2 amplitude. To combine the data from different animals, each individual's N1P2 values were calculated as a percentage of the mean N1P2 value from that animal during the 5 minutes in control conditions (100% level). Single dots with error bars in the main figure represent the mean ± SEM value per 30 sec bin period from n=7 rats. Bar diagram shows the average evoked potential values from n=10 consecutive 30 sec bins during control conditions in air (broken red line), during 5% CO₂ (*p<0.0001 in comparison to control, unpaired t-test) and during recovery in normocapnia (broken green line). Note that the effect of 5% CO₂ on single pulse evoked responses is much smaller than its effect on cortical afterdischarges and myoclonic seizures (see Fig.1).

In all figures depicting EEG data, a downward deflection denotes negativity of the cortical electrode.

Non-human primate epilepsy model

Two macaque monkeys (Macaca nemestrina, weight 2-3 kg) were used, conforming to a Duke University IACUC approved protocol (#A020). Details on the treatment and surgical preparation of macaques for cortical AD stimulation and EEG recording have been described (Haglund et al., 1993; Haglund and Hochman, 2005, 2007). One animal was used for one day of experiments and the second animal for experiments performed on two different days, with 3 months recovery in between. Animals were artificially ventilated on 100% oxygen via intubation (Matrix VBS anesthesia machine and Hallowell EMC model 2002 ventilator) and kept under pentobarbital anesthesia (1-2 mg/kg/hr). Fluid balance was controlled by monitoring intravenous fluid intake every 15 minutes and fluid output (via a Foley catheter) every 30 minutes. Heart rate, blood pressure (Critikon Dinamap blood pressure monitor, model 8100), end-tidal pCO₂, respiration and O₂ levels (Ohmeda 5250 RGM anesthesia monitor; Nellcor NBP-40 for SpO₂) were monitored in line with the EEG throughout experiments.

After craniectomy (25 mm in diameter), the dura was peeled back and surface EEG recordings were made using custom-built strip electrodes (2.5 mm electrode diameter; 1 cm inter-electrode distance) placed over the sensorimotor cortex, with one electrode overlying the hand motor cortex, one over sensory cortex and a reference electrode attached to the skull at the mastoid process.

An acute bicuculline focus was generated in the motor cortex by 0.5 mm² pledgets of gelfoam (Codman & Shurtleff, Randolph, MA, USA) soaked in 100 μM bicuculline (Sigma-Aldrich, St. Louis, Missouri, USA). Pledgets were placed on the cortex for five minutes and refreshed for five minutes every hour. The bicuculline-evoked epileptiform spike activity stabilized after about 1 hour, after which data acquisition started. EEG-signals from motor and sensory cortex electrodes were continuously recorded (1000 Hz sampling, no filters; Axon Instrument's Digidata 1440A system). ADs were evoked by a 4 sec train of 60 Hz biphasic pulses at 4-20 mA using a bipolar stimulating electrode powered by a constant current source (Ojemann Cortical Stimulator, Integra Life Sciences Corporation, NJ, U.S.A.) placed on the sensory cortex. The AD threshold was determined by stimulating at the lowest level that reliably triggered AD activity three times in a row and thresholds remained constant throughout recordings. AD activity was reliably elicited on the sensory cortex during the same experiments in which a bicuculline focus was created on the motor cortex. 4 AD trials were conducted for the two animals, whereby 2 AD trials were carried out on one experimental day with a 20-30 min interval in between. In each AD trial, 3-8 AD stimulations (at 1 minute intervals) were made during the control condition, during hypercapnia and during recovery. Hypoventilation was induced by lowering respiration frequency by 10% and lowering tidal volume, to adjust endtidal pCO₂. End-tidal pCO₂ level was increased from its control level of 37 mmHg to 44-45 mmHg by hypoventilation starting about 10 min before the 2^nd series of AD stimulations, and returned to normocapnia after 8-12 minutes, 5-10 min before the 3^rd stimulation series during recovery. The time needed for reaching the peak level of hypercapnia after the start of hypoventilation was 4 to 6 minutes. Oxygen saturation was measured from the tongue using a tongue-sensor (Nellcor Pulse Oximeter) and was continuously monitored to maintain at constant 98-100% saturation throughout the experiments.

Data were analyzed off-line as described previously (Haglund and Hochman, 2005) with custom-designed software using Python. The bicuculline-induced spike activity was quantified from the motor cortex recording for the first hypoventilation trial on a day of experiments (n=3 trials in 2 animals). The effect of hypoventilation on bicuculline-induced spikes was analyzed by comparing the peak response during high pCO₂ to the mean value during the 5 minutes control period before the start of hypoventilation. In the bicuculline experiments, recovery from hypocapnia was followed for a period of 20-35 minutes after return to normocapnia. Due to technical reasons, one of these three recordings had to be terminated before the recovery phase.

Human epilepsy patients

Subjects were seven patients (average age 32 years, range: 14-52 years; 3 males, 4 females) with drug-resistant partial epilepsy, who were examined with long-term video-EEG recording during pre-surgical evaluation. Details on the patients, including information on medication and seizure times, are given in the Supplementary Information. EEG recordings were obtained using scalp electrodes (n=4 patients; standard International 10/20 positioning system; 32 bipolar channels for EEG registration and analysis) or platinum subdural grid electrodes (n=3 patients). Synchronized video-EEG was recorded and reviewed with the EEG pass-band at 0.5-40 Hz. Seizure behavior included an initial psychomotor or focal motor phase followed by a generalized tonic-clonic activity. One patient (patient 5, see Supplementary information) had only psychomotor semiology. For this patient, EEG was recorded from the left and right frontotemporal region via subdural electrodes and seizure activity was observed to spread bilaterally. The time interval between the seizures included for analysis (see below) varied from 1 hour to 15 hours, with a mean 6.8 ± 1.2 hours (see Supplementary Information for details). Our study had to comply with the local requirements of the pre-surgical clinical workup. In our unit, the number and length of seizures is limited to the minimum that is necessary for a reliable surgical planning, which limits the number of seizures available for our study in each patient. Since it is critical to localize the seizure onset zones, we could start the CO₂ application only after the seizure had generalized (six patients) or had spread to the contralateral hemisphere (patient 5 with only psychomotor semiology) as detected by real-time assessment of the patient's seizure semiology and by the EEG recording monitor in the patient room. Immediately after seizure onset, two technicians entered the room, and CO₂ was applied using a mask with bag connected to a gas bottle containing a mixture of 5% CO₂ in 95% oxygen (medical carbogen; Aga, Finland). CO₂ was administered immediately after seizure generalization (or spreading, in the case of patient 5, see above), typically within one minute after seizure onset, and continued until electrographic seizure termination. Notably, the maximum duration of CO₂ administration was less than 1.5 min (range 20-80 sec). Inhalation flow rate was adjusted to give an excess of carbogen to minimize leakage of room air into the mask. Every seizure was carefully reviewed from the video to confirm the technical success of CO₂ administration. One case with obvious leakage (due to excessive seizure movements) was excluded. The time point of electrographic seizure onset and termination of the seizures from each patient were determined by three clinicians who were blind to the paradigm used. Seizures with uncertain EEG onset and/or termination were excluded from the present analyses. Control seizures were generalized (six patients) or psychomotor (one patient, see above) seizures with comparable semiology and ictal evolution recorded from the same patients. Three control seizures from 3 patients were excluded since no generalization (or, in the case of subdural recordings, significant spreading across the grid) occurred. For 2 of the 7 patients, no control seizures (based on the criteria above) were available. The seizures (n=4) of these patients during which CO₂ was applied were included only for the correlation analysis regarding the effect of the delay to CO₂ application (see Fig. 4B). In total, 7 seizures from 5 patients were included as control seizures and 13 seizures from 7 patients were included for examining the effect of CO₂.

Fig. 4.

Application of carbogen containing 5% CO₂ results in earlier seizure termination in patients with drug-resistant complex partial epilepsy.

A. Example EEG recordings during generalized seizures from 2 patients (scalp EEG, top, and subdural EEG, bottom; overlay of all channels), comparing a control seizure without CO₂ intervention with a seizure from the same patient where 5% CO₂ was applied via a face-mask within about 1 minute after seizure-onset. B. Correlation between the delay to CO₂ application after seizure onset and seizure duration. Data from 13 seizures with application of CO₂, n=7 patients. Spearman r=0.747. C. Seizure durations within the same patient (n=5 patients; 3 scalp and 2 subdural EEG recordings) for seizures without (control, n=7 seizures) and with application of 5% CO₂ (CO₂, n=9 seizures, from the same 5 patients). Application of CO₂ resulted in a significantly shorter seizure duration than what was observed under control conditions (p=0.027 unpaired t-test). CO₂ was applied after seizure generalization, on the average after 40 ± 5 sec from seizure onset (shaded transition from grey-blue); dashed line shows the mean time point used for comparison of seizure duration taking into account the ∼40 sec delay in CO₂ application (see text). See Supplementary Information for further details.

For 4 of the 7 patients, seizures occurred after 1 or more days without medication; see Supplementary Information for further details. All tests were conducted with the understanding and the written consent obtained from the patients and were approved by the Ethics Committee of Helsinki University Central Hospital.

Statistics

Data are represented as mean ± SEM. One-tailed probabilities were used for all statistical tests in our study since, in light of available evidence (see Introduction and Discussion), a plausible hypothesis is that 5% CO₂ suppresses seizures and cortical excitability compared to normocapnic control conditions. For rat AD experiments, responses in the experimental series (CO₂ or acetazolamide) were compared with responses to the corresponding stimulations in the control series, and additional repeated measures analysis of variance (ANOVA) was performed with subsequent pair wise comparison by Bonferroni's test for all AD responses. For the human data, a possible effect of the recording mode on seizure duration was analyzed using a 2-way ANOVA, with the recording type (scalp vs. subdural) as the additional variable. Seizure data from patients with scalp and with subdural electrodes were pooled (see Results) for the data presented in Fig. 4B (scalp: n=8 seizures from 4 patients; subdural: n=5 seizures, 3 patients) and Fig. 4C (control, scalp: n=3 seizures, 3 patients; control subdural: n=4 seizures, 2 patients; CO₂, scalp: n=6 seizures, 3 patients; CO₂ subdural: n=3 seizures, 2 patients).

Results

5% CO₂ suppresses epileptiform afterdischarges and myoclonic seizures in freely behaving rats

ADs were elicited in unanesthetized rats by stimulation of the sensorimotor cortex. ADs consisted of a spike-and-wave pattern seen bilaterally in the EEG (Fig. 1A, B) that was accompanied behaviorally by myclonic seizures (Fig. 1C, D). The AD duration was reduced by 76.5 ± 10% in the presence of 5% CO₂ (Fig. 1A, C; n=8 rats; 8.6 ± 2.3 sec in air vs 2.0 ± 0.9 sec in 5% CO₂; p=0.013, paired-test). In addition to reduced cortical epileptic activity, behavioral seizures were less severe in the presence of 5% CO₂ (Fig. 1C; p=0.016, Wilcoxon matched-pairs signed rank test). Application of 10% instead of 5% CO₂ resulted in a comparable reduction in AD duration (Fig. 1B, D; n=8 rats; p=0.0085, 1-tailed paired t-test; p=0.085, unpaired t-test for comparison between the 5% and 10% CO₂ effect) and intensity of behavioral seizures (Fig. 1D; p=0.0078, Wilcoxon matched-pairs signed rank test). The CO₂ effects on AD duration were found to be significant also using ANOVA followed by a Bonferroni test with multiple comparison among the different AD stimulations (p=0.048 for 5% and p=0.0072 for 10% CO₂).

Fig. 1.

5% CO₂ strongly suppresses cortical afterdischarges and associated myoclonic convulsions in freely behaving rats A, B. Epidural EEG recordings show the cortical AD response in a freely behaving awake rat breathing normal air (left panels in A, B) and 5% or 10% CO₂ in artificial air (right panels in A,B). LS: left somatosensory; LP: left parietal, LO: left occipital; RO: right occipital cortex. C. ADs were reduced by about 75% in the presence of 5% CO₂ in comparison to control stimulations in air (* p=0.013, paired t-test; n=8 rats). In addition, 5% CO₂ strongly reduced the intensity of behavioral seizures accompanying the ADs (*p=0.016, 1-tailed Wilcoxon matched-pairs signed rank test). D. Effect of 10% CO₂ on AD duration (*p=0.0085, paired t-test; n=8 rats) and the associated seizure behavior (*p=0.0078, Wilcoxon matched-pairs signed rank test). Separate animals were used in the tests of 5 and 10% CO₂, with one application of CO₂ tested per group of rats.

During recovery in normal air, AD durations and associated convulsions were similar to those observed under control conditions in air. Fig. 1C and D show the responses for the first stimulation during the recovery, which were not significantly different from control air responses (p=0.36 for the AD durations after 5% CO₂ and p=0.35 for the AD durations after 10% CO₂ exposure, respectively, in comparison to responses in the control sessions in air; paired t-test). An additional stimulation 20 minutes after the first stimulation during recovery evoked ADs of similar duration (7.3 ± 1.8 sec for the last vs 7.3 ± 2.3 sec for the first recovery response after 5% CO₂; 8.7 ± 2.4 sec for the last vs 7.9 ± 1.6 sec for the first recovery response after 10% CO₂). Separate groups of animals were used for each condition (i.e. 5% and 10% CO₂), with a single application of carbogen.

Systemic administration of the carbonic anhydrase inhibitor acetazolamide (50 mg/kg) in a separate group of animals (n=10 rats) resulted in a significant, but smaller, reduction of AD duration than observed for 5% CO₂: a reduction of AD duration by 50-55%. The effect of acetazolamide was evident for the 3rd, 4rd and 5th stimulation, at 35 minutes, 55 min and 75 min after drug injection (p-values for comparison to corresponding responses in the vehicle control session: p=0.011, p=0.0051; p=0.0063, paired t-test) and was strongest at 55 min after acetazolamide injection (55.6 ± 9.1 % reduction, compared to 100 ± 153% with vehicle, p=0.0013, repeated measures ANOVA followed by Bonferroni test).

To gain insight into the time-course of the CO₂ effect on cortical excitability, we administered 5% CO₂ while recording single evoked potentials in 7 rats (Tsenov et al., 2009). In the presence of 5% CO₂, the peak-to-peak amplitude (N1P2) of cortical potentials was reduced by 16% ± 2.0 % from the values during control conditions (100 ± 1.5%; p=2.7·10^-8, unpaired t-test). The effect of 5% CO₂ was visible within 30 seconds after start of the carbogen administration (Fig. 2). Taking into account the time needed for the gas mixture in the recording chamber to reach its final concentration, which was confirmed by capnographic measurements to take about 15 sec (see Methods), these data demonstrate a rapid suppression of cortical excitability by 5% CO₂. Notably, the effect was much smaller than that observed above for evoked cortical ADs. Taken together, these data in unanesthetized rats showed that elevating CO₂ to 5% has a strong suppressing effect on epileptiform ADs, but much less on individual cortical potentials.

Mild hypercapnia blocks cortical afterdischarges in macaques

ADs were evoked in 4 separate trials in 2 macaque monkeys by stimulation of the sensorimotor cortex. End-tidal partial CO₂ pressure was increased by hypoventilation from its control level of 37 mmHg to 44-45 mmHg, an increase similar to that observed in humans inhaling 5% CO₂ (Voipio et al., 2003). Strikingly, in all 4 trials, the slightly elevated pCO₂ strongly suppressed and in many cases completely blocked the electrically-induced ADs (Fig. 3A). The hand movements that are typically associated with the ADs in normocapnia were also blocked. Full recovery of evoked AD activity and associated convulsive movements was evident after return to normocapnia (Fig. 3A). Next, we studied the effect of hypoventilation on epileptiform spike activity, generated by focal bicuculline application on hand motor cortex. Increasing pCO₂ caused a rapid but slight reduction in spike amplitude, with a peak reduction by 23.1 ± 5.6 % from mean control values observed 2-4 minutes after constant 44-45 mmHg pCO₂ was reached (710 ± 70 uV during the control condition in normocapnia vs 570 ± 110 uV during the peak response at high pCO₂; p=0.038, paired t-test; n=3 trials; 2 animals; Fig. 3B). Spike frequency was reduced by 15.2 ± 5.1 % (58 ± 1 spikes/min during control conditions vs 50 ± 3 spikes/min at the peak response during increased pCO₂; p=0.048, paired t-test; n=3 trials; 2 animals).

Fig. 3.

A. Cortical ADs evoked in the sensorimotor cortex in macaque monkeys are strongly suppressed or completely blocked by increased levels of pCO₂. Surface EEG recordings show an evoked AD during control conditions (top trace; end-tidal pCO₂ 37 mmHg), a complete block of AD activity during mild hypercapnia at a level of 44mmHg pCO₂ (middle trace; 8 minutes after end-tidal pCO2 was increased to 44 mmHg); and a full recovery of AD activity 5 minutes after return to normocapnia (bottom trace). Hand movements observed during ADs were also suppressed during the increase in pCO₂. Note ongoing spike activity generated by an acute bicuculline focus about 1 cm away which is not completely blocked in high pCO₂ (details in B). Bar diagram on the right shows the significant reduction in AD duration during increased pCO₂ by about 70% from that observed during control conditions, and the full recovery (* p=0.0085 paired t-test; n=4 trials, 2 macaques). B. Bicuculline-induced epileptiform activity is reduced in amplitude when pCO₂ is increased to 44-45 mmHg. Example EEG recording from the hand motor cortex nearby the bicuculline focus showing spike activity during control normocapnia (37 mmHg pCO₂) and, with slightly reduced amplitude, during the 5^th minute in high CO₂ (45 mmHg). Bar diagram shows a reduction in spike amplitude by about 25% of control during increased pCO₂ (* p=0.038, paired t-test; n=3 trials, 2 macaques). Recovery of spike amplitude was recorded only for 2 of the 3 bicuculline trials.

Inhalation of medical carbogen causes termination of generalized seizures in epilepsy patients

The observations in rodents and macaques demonstrated that a level of 5% CO₂ effectively suppresses cortical epileptiform activity and associated behavioral convulsions. Next, we applied medical carbogen (5% CO₂ + 95% O₂) using a respiratory facemask shortly after electrographic onset of 13 spontaneous seizures in 7 partial epilepsy patients. For 6 of the 7 patients, EEG seizures were accompanied by generalized tonic-clonic behavior. One patient without generalized seizure activity had psychomotor seizures that spread bilaterally. In all patients seizures were accompanied by loss of contact. Seven control seizures from the same patients were selected on the basis of comparable generalization times or significant spreading pattern across the subdural grid. For two patients no comparable control seizures were available (see Methods and Supplementary Information for details). The 4 CO₂-treated seizures from these patients were included only for analysis of the effect of the delay to CO₂ application on seizure duration (see below).

Fig. 4A shows a comparison of EEG recordings from patients during a generalized seizure with and without CO₂ administration. Upon administration of 5% CO₂, seizures terminated within 16-77 seconds (mean 40 ± 6 sec; n=13 seizures, n=7 patients). The delay to CO₂ application varied from 23 to 74 seconds (mean 40 ± 5 sec) after electrographic seizure onset. Notably, longer seizure duration correlated significantly with a longer delay to CO₂ application (Fig. 4B; Spearman r=0.747, p=0.0034). Seizures during which CO₂ was applied were significantly shorter in duration than control seizures recorded from the same patients (Fig.4C; p=0.027, unpaired t-test; 109 ± 13 sec for n=7 control seizures vs 76 ± 9 sec for n=9 CO₂-seizures, n=5 patients). With CO₂ administration, the duration of seizures was reduced by 29.7± 7.6 % compared to control seizures recorded in the same patient (p=0.024, unpaired t-test).

An initial glance at the primary data suggested (as perhaps might have been expected) that seizure durations were longer with subdural than with scalp recordings. However, there was a significant overlap for scalp and subdural seizure durations within both the control and CO₂-treated groups. In accordance with this, when tested using 2-way ANOVA (see Methods), the seizure duration data as a whole did not show statistical significance between the two types of recordings (p = 0.08). Hence, the data from scalp and subdural recordings were pooled for the analyses shown in Fig. 4.

As noted, the delay to CO₂ application was about 40 sec after the start of a seizure (see also Fig. 4C). Hence, a valid estimate of the effect of 5% CO₂ is given by comparing seizure durations based on the time point of onset of CO₂ administration (see dashed line in Fig. 4C). Measured from the time point of CO₂ application, seizure duration was 32.1 ± 6.6 sec when CO₂ was administered compared to 65.4 ± 12.3 sec under control conditions. This indicates a 50% reduction in seizure duration by application of 5% CO₂ (p=0.012, unpaired t-test; n=9 CO₂-seizures and n=7 control seizures, n=5 patients.

Discussion

The present work shows that 5% CO₂, present in standard medical carbogen, has a potent effect on cortical epileptiform activity and convulsions both in animal epilepsy models and in humans with drug-resistant partial epilepsy. Importantly, we observed that the 5% CO₂ level proved effective, and prolonged administration is not needed: seizures in the patients observed in our study terminated typically within 1 minute after carbogen was applied. Unfortunately, the constraints imposed by the presurgical evaluation of the patients made it possible to apply carbogen only after seizure generalization (see Methods). It can be safely predicted on the basis of our data that 5% CO₂ will have a stronger anticonvulsant effect in future clinical trials and therapeutic interventions if the application can be started immediately after seizure onset.

pCO₂ and pH effects on neuronal excitability

Apart from the suppression of spike-wave EEG activity by 10% CO₂ in patients with absence epilepsy (Lennox et al., 1936), small increases in pCO₂ in healthy human subjects are associated with a suppression of EEG power (Bloch-Salisbury et al., 2000). In contrast, the excitatory effect of a hypocarbic alkalosis is often seen in humans as hyperexcitability associated with hyperventilation (Foerster, 1924; Rosett, 1924; Lennox et al., 1936; Huttunen et al., 1999; Guaranha et al., 2005; Sparing et al., 2007).

The acidosis caused by elevated pCO₂ has a direct suppressing effect on brain excitability in vivo (Balestrino and Somjen, 1988), a conclusion that gains strong support from experiments done on brain slices (see below). Retention of CO₂ and subsequent changes in brain pH through systemic carbonic anhydrase inhibition by acetazolamide may account for its anticonvulsant properties, as indicated by earlier work on acetazolamide in rodent epilepsy models in vivo (Rollins et al., 1970; Anderson et al., 1986). An alternative or additional mechanism to suppress excitability is a direct effect of acetazolamide on intraneuronal carbonic anhydrase (Viitanen et al. 2010). Notably, the suppression of seizure activity by acetazolamide in our rat experiments, at an established dose of 50 mg/kg, was about 25% less than that observed for 5% CO₂.

Studies in which CO₂ was used to suppress neuronal activity in vitro indicated a key role for pH in its effect (Velisek et al., 1994; Lee et al., 1996; Dulla et al., 2005) and showed that effects of CO₂ on cerebral blood flow are not the basis of the anticonvulsant action. This is supported further by several observations of reduced neuronal activity in vitro in response to acidosis (Speckman and Caspers, 1974; Aram and Lodge, 1987; Kaila, 1994; Velisek et al., 1994; Lee et al., 1996; Kaila and Ransom, 1998; Schweitzer et al., 2000; Dulla et al., 2005; Dulla et al., 2009). Possible cellular mechanisms (see Kaila and Ransom, 1998; Chesler, 2003, for review) involve direct effects of pH on voltage and ligand-gated ion channel conductances (Traynelis et al., 1995; Lee et al., 1996; Pasternack et al., 1996; Ziemann et al., 2008) and adenosine signaling (Dulla et al., 2009).

CO₂ levels higher than 5% and/or prolonged application times might have adverse effects due to a rebound alkalosis (‘off’-response) upon return to normocapnia, resulting in increased cortical excitability (Aram and Lodge, 1987; Lee et al., 1996). In the present work, no rebound increase in excitability was observed upon return to normocapnia.

Role of pH in self-termination of seizures: additive effect of exogenous CO₂?

In previous work (Schuchmann et al., 2006), we have shown that experimental febrile seizures in rat pups are caused by a respiratory alkalosis, and that 5% CO₂ blocks these seizures very quickly (in about 20 s) by abolishing the rise in brain pH (see also Schuchmann et al., 2009). In the present study, seizures were not triggered by an alkalosis and hence the effect of 5% CO₂ is attributable to a net acidosis as discussed above. Notably, neuronal activity itself causes changes in pH. During seizures or excessive neuronal activity in animal models in vitro (de Curtis et al., 1998; Xiong et al., 2000; Dulla et al., 2009) and in vivo (Wang and Sonnenschein, 1955; Meyer et al., 1961; Caspers and Speckmann, 1972; Somjen, 1984; Siesjö et al., 1985; Xiong and Stringer, 2000) an acidosis develops, which may contribute to seizure termination. The acidosis can result from metabolic production of CO₂ and lactate, as well as from net entry of acid through ligand-or voltage-gated channels (Kaila, 1994). The proposed effect of acidosis on seizure termination is likely to involve adenosine, levels of which were shown to increase upon decreases in pH during hyperexcitability in vitro (Dulla et al., 2009). That CO₂-production is significant during seizures was shown recently for chronic epilepsy patients, in which a pronounced increase in end-tidal pCO₂ occurred during seizures that could persist for several minutes. The rise in pCO₂ was attributed to seizure-related hypoventilation (Bateman et al., 2008). In this context it is worth noting that changes in ventilation patterns have an effect on seizure frequency (Fried et al., 1990). Observations of increased seizure occurrence associated with slight changes in ambient air pressure (Sirven et al., 2002; Doherty et al., 2007) are consistent with the high sensitivity of seizures to small changes in pCO₂.

It is likely that the administration of CO₂ during seizures exerts a combined seizure-terminating effect together with endogenous increases in neuronal pCO₂, pH and/or adenosine. This could explain why the suppressive effect of 5% CO₂ was much stronger for cortical ADs in both rats and macaques when compared to single cortical spikes. Similarly, in an earlier study in non-human primates, inhalation of 6% CO₂ caused only a slight reduction (about 15%) of spontaneous cortical multi-unit activity and of EEG activity in the beta and gamma ranges (Zappe et al., 2008).

Therapeutic value and clinical safety of 5% CO₂

Our pilot trial carried out in humans supports the idea that application of medical carbogen containing the standard level of 5% CO₂ is effective for the acute treatment of seizures in human patients. In light of our results, the use of CO₂ may turn out to be particularly suitable to treat prolonged seizures in the setting of the emergency room, intensive care unit or ambulance. In addition, since some chronic epilepsy patients can anticipate the start of seizures (Petitmengin et al., 2006; Haut et al., 2007), prophylactic treatment might become an option outside the clinic, provided that professional supervision is available. Optimal concentrations are expected to vary among patients and situations, and the transient nature of CO₂ action has to be taken into account in such studies. Therefore, dose-response data should include information on the timing and duration of carbogen application, preferably with the start of application as soon as possible after seizure onset.

Clinical safety is a major concern in a pioneering test of the present kind. Carbogen containing a level of 5% CO₂ has been used to enhance the effectiveness of radiotherapy for tumour patients (Baddeley et al., 2000; Taylor et al., 2001) and also in other kinds of therapies and clinical trials (Kergoat and Tinjust, 2004; Ashkanian et al., 2009). Little discomfort was reported with patients breathing 5% CO₂ for a couple of minutes (Kergoat and Tinjust, 2004; Ashkanian et al., 2009), which is in line with observations in healthy volunteers (Voipio et al., 2003). However, prolonged application of 5% CO₂ (Ley, 1991) and higher CO₂ percentages (Bailey et al., 2005; Griez et al., 2007) can give rise to symptoms of anxiety. Hence we based our study on using brief periods of application of 5% CO₂.

In the present work the effects of 5% CO₂ on non-epileptic cortical activity were small compared to its effect on seizure activity. Due to the short application time needed, effects on normal brain function and other side effects are expected to be minimal. Our pilot human data, combined with the experimental results from the parallel animal studies, are encouraging in showing that a rapid seizure-suppressing effect can be achieved with 5% CO₂. The low cost of medical carbogen makes the inhalation approach also available for patients in less-developed countries.