Phenobarbital, Midazolam, Bumetanide and neonatal seizures: the devil is in the details!

Summary

Kaila, Löscher and colleagues report that Phenobarbital (PHB) and Midazolam (MDZ) attenuate neonatal seizures following birth asphyxia, but the former only when applied before asphyxia and the latter before or after the triggering insult. In contrast, the NKCC1 chloride importer antagonist Bumetanide (BUM) had no effect whether applied alone or with PHB. The observations are compelling and in accord with earlier studies. Yet, there are several general issues that deserve discussion. What is the clinical relevance of these data and the validity of animal models of encephalopathic seizures? Why is it that although they act on similar targets, these agents have different efficacy? Are both PHB and MDZ actions restricted to GABAergic mechanisms? Why is BUM inefficient in attenuating seizures but capable of reducing the severity of other brain disorders? We suggest that the relative failure of Anti-Epileptic Drugs (AEDs) to treat this severe life-threatening condition is in part explicable by the recurrent seizures that shift the polarity of GABA, thereby counteracting their effects on their target. AEDs might be efficient after a few seizures but not recurrent ones. In addition, PHB and MDZ actions are not limited to GABA signals. BUM attenuates efficiently Autism symptomatology notably in patients with Tuberous Sclerosis but does not reduce the recurrent seizures, illustrating the uniqueness of epilepsies. Therefore, the efficacy of AEDs to treat babies with encephalopathic seizures will depend on the history and severity of the seizures prior to their administration, challenging a universal common underlying mechanism.

Commentary

The development of novel Anti-Epileptic Drugs (AEDs) is reminiscent of the Russian roulette with its ups and downs. Many promises have had a short lifetime, notably with successful animal models that are not readily translatable to clinical setups. Failure of AEDs are also due to controversial experimental observations, inadequate underlying concepts or a mixture of both. In a recent review, Seyhan reported the difficulty of translating basic science to treatments with over 95% of failures relying on preclinical research (1). These difficulties are particularly numerous when dealing with neonatal epilepsies.

The excellent original research article by Johne et al. vividly illustrates the effects and lack of effects of AEDs on asphyxia-induced neonatal seizures (2). In a new neonatal asphyxia model, the authors report that Phenobarbital (PHB) blocked the seizures when applied prior to the asphyxia episodes but not after, while Midazolam (MDZ) attenuated the seizures when applied before or after. The NKCC1 specific antagonist Bumetanide (BUM) was inefficient when applied before or after the seizures, and did not ameliorate the effects of PHB contrary to earlier studies (3)(4). These observations are in keeping with earlier reports including our own (5)(6)(7). Nevertheless, there are many factors that must be worth considering before accepting or discarding these agents to treat neonatal epilepsies and assigning an exclusive underlying mechanism to their effects. Indeed, the failure of inhibition and the Excitatory/Inhibitory balance stands at the core of several brain pathologies extending from seizures to Autism Spectrum Disorders (ASD), Parkinson Disease, schizophrenia and brain damage (7). Yet, this is validated in some preclinical and clinical conditions, and not in others. Is there something distinctive about seizures and epilepsies? Are neonatal epilepsies unique? Do these 3 companion drugs act similarly on the same mechanisms? Below, we suggest a possible scenario explaining why PHB and MDZ might lead to different outcomes in addition to their effects on GABAergic currents, and why they fail - or worse - generate paradoxical effects. We also discuss briefly the pros and cons of the actions of BUM and the reasons why it is unlikely to efficiently treat neonatal epilepsies.

GABA, chloride and seizures: timing, frequency and chronicity of recurrent seizures are crucial

A plethora of unequalled complex mechanisms underlie the actions of GABA. GABAergic inhibition is particularly activity-dependent and can even reverse polarity (5)(8)(9)(10)(11). This is unique to GABAergic (and glycinergic) currents. Thus, by measuring intracellular chloride concentration ([Cl⁻]_i) levels with perforated patch electrodes in pyramidal neurons of the hippocampus, we determined the duration required to recuperate control values of [Cl⁻]_i levels after exciting the neuron. Neonatal neurons require a longer period of time than adult ones, and even more so for epileptic neurons, attesting to the long term effects of recurrent seizures on GABA polarity (Figure 1) (12). In these conditions, recurrent, but not single propagating seizures, altered quasi permanently [Cl⁻]_i levels and GABA polarity (13)(14). This is a sort of LTP, whereby seizures beget seizures by altering progressively in a more permanent manner the capacity of neurons to regulate their [Cl⁻]_i levels (12). AEDs can block the first seizure, but not after many epileptiform events (Figure 2). Therefore the efficacy of AEDs depends on its timing of administeration. BUM did not prevent the formation of a mirror focus by seizures, but it attenuated the severity of seizures in the mirror focus generated in the isolated hippocampus by the propagation of recurrent seizures (6). Therefore, the severity and frequency of seizures must be incorporated in any analysis of the efficacy or lack of efficacy of AEDs. An AED can attenuate seizures on one child having few recurrent seizures but not another one with many frequent seizures. Collectively, these observations underlie the heterogeneity of AEDs efficacy.

Figure 1.

Neurons were recorded with perforated patch clamp electrodes and puff applications of GABA were performed. Initially (left side), GABA evoked no current as the holding membrane potential was adapted to the chloride equilibrium potential. Large outward GABA currents are generated when the neuron is strongly depolarized. GABA evoked a large inward current subsequently because of this strong influx of chloride, and the time required to remove excess chloride and return to control values provides an indication of the efficacy of chloride removal. Note that the kinetics is much faster in control neurons (A) than in epileptic neurons recorded in a mirror focus MF (B). Adapted from Y. Ben-Ari, 2012, Epilepsia (12).

Figure 2.

Phenobarbital (PHB) attenuates early seizures but aggravates late ones. Here, kainate was applied to one hippocampus to generate interictal and ictal discharges (dark trace) that propagate to the contralateral hippocampus. PHB application to contralateral hippocampus (red trace) efficiently blocked the first paroxysmal activity (Ia), but aggravated them in the 15^th discharge (Ib). Right side: the different time frequency power plots after 1 (above) and 15 (below) propagated epileptiform activity in the presence of PHB. Adapted from Nardou et al., 2011 (5).

PHB, MDZ and neonatal seizures: less than ideal

As stressed by Johne et al. (2), Phenobarbital (PHB) has toxic effects and at best is efficient to treat seizures in half of the neonates with various forms of epilepsies, yet it is the first therapy in line to treat seizures. MDZ, a second line treatment, is a largely used AED with limited efficacy and side effects, in particular as far as encephalopathic seizures are concerned. The classical mechanism thought to underlie the effects of MDZ and PHB is a potentiation of GABAergic inhibition and the large Cl⁻ influx. However, several observations suggest that the actions of these AEDs are not solely mediated by GABAergic mechanisms. In the triple chamber preparation, where two intact neonatal hippocampi are separated by a chamber allowing to perfuse each with a different liquid (6)(15), PHB but not MDZ reduced interictal events generated in the presence of GABA and NMDA receptor antagonists (16). In fact, MDZ enhanced GABAergic excitatory events and aggravated the epileptiform activity generated in a mirror focus (16). Also, PHB but not MDZ reduced the frequency of AMPA receptor mediated EPSCs in the presence of GABA receptor blockers suggesting a direct effect on Glutamatergic synaptic currents (16). In the presence of GABA signaling blockers, another Benzodiazepine (Zolpidem) reduced significantly calcium currents (17). These in vitro experiments confirm the heterogeneity of actions of PHB and MDZ and their different effects on epileptiform activity. PHB or MDZ might also exert different actions on various GABAergic hub interneurons such as those reported to exert a long distal control of septo-hippocampal seizures (18). We do not know whether hyperactivity episodes impact different neuronal types that might be differently susceptible to the actions of PHB or MDZ. In clinical conditions, hypothermia, which is the most effective treatment of neonatal encephalopathy, alters clearance of PHB and MDZ to different degrees, shortening their efficacy (19). MDZ has hypotensive effects that alters pharmacokinetics and has poor efficacy as first or second line AEDs after PHB (20). This contrasts with the efficacy of MDZ administered before or after the insult in rodents (2). Clearly, comparing the efficacy of PHB and MDZ solely on the basis of their GABA actions does not provide a full image of their effects (also see below).

The Bumetanide story: successes and failures!

Since the discovery that high neuronal [Cl⁻]_i levels occur in many pathological conditions, a large number of studies have tested the actions of BUM (or KCC2 antagonists) in animal models. In epilepsies, controversial studies have been published with attenuations, no effects or even aggravation, the activity-dependence of GABA actions hampering the possible AEDs action of BUM (21)(22). However, the treatment of Autism Spectrum Disorders (ASD) in independent pilot case trials and double-blind randomized phase 2 trials attest to its efficacy in attenuating social interactions (23)(24)(25)(26). Visual eye-tracking and functional brain imaging attest also to the efficacy of BUM in ASD (27)(28). In addition, BUM was recently shown to attenuate autism syndrome severity in Tuberous Sclerosis, but not the seizures often present in these children (29). This illustrates the difficulty of BUM to act as an AED, while in the same patients it does reduce ASD symptomatology. Pilot trials also suggest that BUM might be efficient in Fragile X syndrome (30), schizophrenia (31) and Parkinson Disease (PD) (32). In keeping with this, GABAergic inhibition fails in striatal cholinergic interneurons in a model of PD (33). Collectively, these observations suggest that attenuating high [Cl⁻]_i levels in brain disorders remain an interesting option without being generalized to the variety of epilepsies.

BUM has been shown to attenuate seizures in some pathological models but not others (34)(35). As discussed by the authors, BUM showed some efficacy to attenuate seizures in 2 studies of post asphyxia seizures in neonatal rodents but not in another one. Although Johne et al. (2) and Cleary et al. (36) used experimental conditions with very minor differences (15 and 30 mins asphyxia), they observed opposite results, negative in the former positive in the latter. The reasons for the discrepancy remain unclear but attest to the complexity of the effects of seizures and AEDs. Kang et al. (37) used anesthesia to perform carotid ligation followed by the administration of PHB (1 hour) or BUM (2 hours). PHB attenuated seizures in postnatal day 10 (P10) and P12 but not P5 rodents. However, the effects of isoflurane anesthesia on electrical activity should not be underestimated in particular with a short delay (38)(39). Interestingly, the lower the seizure burden the better the anti-seizure efficacy of PHB, again stressing the activity-dependence effect of PHB. Wang et al. (40), used a different paradigm. Rats were subjected to hypoxia-induced neonatal seizures at P10, and treated for 3 weeks with intraperitoneal injections of BUM at a dose of 0.5 mg/kg. These repeated long-lasting administration of BUM attenuated the seizures, but also cell proliferation and mossy fiber sprouting (40). These more chronic experiments are therefore not comparable with the earlier ones with their short duration BUM administration. Interestingly, in children with ASD, the actions of BUM are not immediate (24). Therefore, in a clinical set up of recurrent life-threatening conditions, BUM is unlikely to exert an important AED action because of these limitations.

The authors dwell substantially on the NEMO trial to illustrate the lack of efficacy of BUM to attenuate encephalopathic epilepsies in babies (41)(42). In the NEMO trial, the trial was stopped because of ototoxicity suggesting severe side effects. Hearing loss was observed in 3 out of 11 surviving babies. However, it bears stressing that the babies received concomitantly large doses of aminoglycosides also known to have ototoxic actions. Indeed, NKCC1 is involved in the development of the auditory brainstem and in the production of K⁺-rich endolymph, underlying the sensitivity to BUM (43)(44)(45)(46), but the concomitant application of aminoglycosides strongly aggravates ototoxicity (47). The actions of BUM are potentially toxic at an early stage prior to the shift of the GABA Excitatory/Inhibitory shift (48), but BUM has been used in babies previously without ototoxicity (49). In the NEMO trial, Pressler and colleagues have stressed at length the difficulty to determine benefit/risk in life threatening conditions (41). In addition, in the NEMO trial, BUM was only administered after a loading dose of PHB had failed to attenuate the seizures. The impossibility to use BUM as a first intention treatment and the many seizures occurring before the treatment necessarily hampers the conclusions on its efficacy (also see above). There are also several obvious reasons why the NEMO trial is irrelevant to the use of BUM to treat brain disorders. In the NEMO trial BUM was injected intra-venously at very high doses that might indeed be toxic and are certainly not used in other clinical situations. Therefore, we agree that BUM is an unlikely adequate AED to treat encephalopathic seizures but this limited trial does not preclude its utility in other disorders including other types of epilepsies.

In addition, because of the limited amount of BUM that penetrates the brain through the blood-brain barrier (BBB), Kaila, Löscher and colleagues (2)(50)(51) have repeatedly challenged its efficacy to treat brain disorders. Indeed, studies have shown that the levels of BUM found in the brain of rodents are below the amounts needed to act on brain NKCC1. However, the main reason for this poor brain availability is the fact that ~97% of BUM binds to serum proteins, e.g. albumin (52). The stoichiometric binding constant for BUM is 2.29 x 10⁴ M⁻¹ in neonatal blood. BUM competes with bilirubin for human serum albumin binding, having a displacement constant (K_Disp) of 6.2 × 10³ M⁻¹ when measured by the peroxidase method. This displacement effect is also observed using pooled umbilical cord serum and pooled adult serum by employing a dialysis rate method. BUM competes to a lesser degree with Diazepam binding to human serum albumin. No competition with Diazepam occurs using umbilical cord or adult serum. Pharmacologic concentrations of BUM would not significantly affect bilirubin-albumin binding and should not increase the risk of bilirubin encephalopathy in newborn infants. In addition, brain levels are, as expected, higher after the asphyxia insult (36). An indirect action first on non-BBB protected brain areas like the centers involved in hormonal regulation have been suggested as a possible explanation to this conundrum (53). Yet, alterations of BBB in humans in pathological conditions cannot be excluded.

It is difficult to explain how BUM could in experimental conditions attenuate so many different pathological insults extending from cerebro-vascular infarcts, PD, chronic pain, ASD, genetic forms of developmental disorders, Alzheimer and even brain tumors without a central action (7). Extensive evidence of alterations of vasculature and BBB have been observed notably in ASD, Fragile X syndrome and schizophrenia with abnormalities of brain vasculature, increased levels of markers associated with endothelial cells and pericytes (reviewed in (54)). Therefore, rodent data are not readily applicable to humans and more measures of central BUM levels being required. BUM remains a promising treatment for many brain and peripheral disorders.

Conclusion

To conclude, neonatal asphyxia encephalopathy associated with seizures are difficult to treat and neither PHB nor MDZ are ideal AEDs. BUM is unlikely to be efficient primarily because of its slow actions, requiring days to exert an effect, and not its postulated ototoxicity. Furthermore, the actions of BUM, like that of PHB or MDZ, are hampered by the augmented neuronal [Cl⁻]_i levels and shift of GABA polarity produced by recurrent seizures. The efficacy of their actions is therefore strongly dependent on the severity and frequency of seizures taking place before their administration.