Targeting neuronal homeostasis to prevent seizures

Manipulating neuronal homeostasis, which enables neurons to regulate their intrinsic excitability, offers an attractive opportunity to prevent seizures. However, no anticonvulsant compounds have yet been reported that directly manipulate neuronal homeostasis. Here, we describe a novel class of anticonvulsant compounds, based on 4-tert-butyl-benzaldehyde (4-TBB), with a mode-of-action that includes increased expression of the homeostatic regulator Pumilio (Pum). In Drosophila and mouse we use a pentylenetetrazole (PTZ) induced seizure model, and an electrically induced seizure model for refractory seizures to evaluate anticonvulsant efficacy. The pyrazole analogue (RAB216) demonstrates best efficacy, protecting 50% of mice from PTZ-induced seizure. Knock-down of Pum, in Drosophila, blocks anticonvulsive effects, whilst analysis of validated Pum targets show significant reductions following exposure of mouse brain to 4-TBB. This study provides proof-of-principle that anticonvulsant effects can be achieved through regulation of neuronal homeostasis and identifies a chemical lead compound for future development.


Introduction
Neuronal homeostasis provides an attractive target to achieve therapeutic control of epilepsy. This is because homeostatic mechanisms ostensibly oppose extremes of neuronal activity that are associated with seizures. By maintaining neuronal activity patterns at physiologically relevant 'setpoints', neuronal homeostasis acts to ensure stability of both neuron and network function across the life-course 1 . However, to date, neuronal homeostasis has not been specifically targeted for clinical benefit.
Pumilio (Pum) homeostatically maintains action potential firing rates within a set-range 2 . A translational repressor, Pum binds mRNA transcripts and reduces de novo protein synthesis, with increased Pum expression occurring in neurons exposed to increased synaptic excitation.
Conversely, as synaptic excitations fall, Pum expression is reduced 3 . The 3'-UTR of Pum-regulated transcripts usually contain one or more copies of a Pum-Response Element (PRE: UGUANAUA, where N is A, C, G or U) 4 . Analysis of both Drosophila and mammalian transcriptomes identifies more than 1000 transcripts that contain one or more PREs, consistent with a broad regulatory role 5 .
However, Pum requires additional co-factors (including Nanos and Brain-tumor) and the actual effect of Pum is likely dictated by both the number and proximity of these additional binding elements, in addition to the number of PREs 6,7 . In mammals, regulated transcripts that have potential to influence neuron activity include Nav1.6 (SCN8A) 2 and GLUR2 (AMPA receptor) 8 .
Pum-dependent homeostatic translational repression of Nav1.6, in rat cortical pyramidal neurons, reduces the amplitude of expressed voltage-gated Na + current (INa) and lowers action potential firing frequency 2 . Down-regulation of AMPA receptor expression may also be expected to be antiepileptic, evidenced by the anti-epileptic compound, perampanel, which is an allosteric antagonist of AMPA receptors 9 .
Whilst Drosophila has one pum gene, mammals express two highly similar variants (pum 1 & 2) that are co-expressed, and which bind identical RNA motifs and, thus, appear to act redundantly [10][11][12] . Seizure occurrence could reflect reduced homeostatic capability and it is significant that recent studies suggest reduced Pum contributes to epilepsy. Specifically, i) pum1 or 2 haploinsufficiency is associated with spontaneous seizures in mice [13][14][15] , ii) Pum2 expression is reduced in human patients suffering temporal lobe epilepsy and in rat hippocampus following pilocarpine-induced seizure 16 , and iii) dpum expression is reduced in Drosophila genetic seizure mutations 17 . In the latter, transgenic up-regulation of dPum is potently anticonvulsant in these same Drosophila mutations 17 .
Based on a screen to identify chemicals that increase expression and/or stability of dPum, we identified avobenzone, which secondary screens show is anticonvulsant in seizure-sensitive Drosophila 17 . However, the physiochemical properties of avobenzone are not compatible with clinical use. Thus, in this study, we report the identification of an avobenzone analogue, 4-tertbutyl-benzaldehyde (4-TBB), that is anticonvulsant and has properties more consistent with clinically active compounds. We show that 4-TBB and analogues (specifically the pyrazole RAB216) are active against a range of Drosophila seizure mutants and, significantly, reduce severity of both pentylenetetrazol (PTZ)-and pharmacoresistant electrically (6Hz)-induced seizures in mouse. Reduction of seizures, in fly and mouse, is accompanied by increased expression of Pum.
We further report down-regulation of known Pum targets following exposure of mouse brain to 4-TBB thus validating proposed mode-of-action.

4-TBB suppresses seizure behaviour in Drosophila
Single Drosophila gene mutations increase seizure-like activity in response to electrical shock 18 . In a prior study, we identified avobenzone to both increase dpum expression and reduce seizure severity in Drosophila 17 . Avobenzone is poorly soluble and therefore we identified a breakdown product, 4-TBB to be a better candidate for analysis of mode-of-action. We tested the anticonvulsant activity of 4-TBB in three diverse seizure mutants to demonstrate wide applicability.
The expression of dpum is reduced in para bss mutants and, moreover, increasing dpum expression in this mutant background is sufficient to suppress seizure activity in response to electroshock 17 . To determine whether the anticonvulsant effect of 4-TBB is associated with up-regulation of dpum, we used a dpum-minimal promoter construct upstream of GAL4 (dpum-GAL4) to drive expression of UAS-luciferase (UAS-luc) 19 . Exposure of dpum-GAL4>UAS-Luc flies to 4-TBB (1.2 mM in food) resulted in a significant increase in luciferase activity (2.7 ± 1.5 fold change, n = 5, p = 0.03, vehicle control set as 1) (Fig. 1B). We adopted this approach because available anti-Pum antibodies (designed to rodent Pum1 and Pum2) do not work well in Drosophila. We also observed a significant increase in dpum transcript abundance, measured by QRT-PCR, of ~60% (1.6 ± 0.3 fold change, n = 5, p = 0.002, vehicle control set as 1) following exposure to 4-TBB (Fig. 1C). Finally, we find that the anticonvulsive activity of 4-TBB is significantly diminished when dpum expression is reduced in the CNS, via targeted expression of an RNAi construct (Fig. 1D). PHE, which has a different mode of action 20 , remains active under these conditions. We conclude that the mode-ofaction of 4-TBB requires the presence of dPum and increases expression of this homeostatic regulator.

4-TBB reduces epileptiform activity in mouse hippocampal culture slices
Incubation of acutely-harvested mouse brain slice with 4-TBB (1.2mM) for 2h was sufficient to produce a significant increase in Pum2 expression (Pum1 not measured), as determined by Western Blot (2.5 ± 0.3 fold compared to vehicle-treated controls which were the corresponding slice taken from the opposite hemisphere incubated in vehicle only, set at 1, n = 5, p = 0.002, data not shown).

4-TBB reduces induced PTZ-induced seizure in mice
Mice were exposed to 4-TBB (800 mg/kg, s.c. injection) or saline vehicle (CTRL), once per day for 3 days. We observed no overt change in behaviour, nor weight loss, during the test-period. Four hours after the last injection on day 3, seizures were induced by PTZ (60 mg/kg, s.c. injection).

Identification of a more potent 4-TBB analogue
The above data shows proof-of-principle that a compound that effects increased Pum activity has significant potential as an anticonvulsant. However, the active concentration of 4-TBB required for significant effect, in the in vivo mouse PTZ-induced seizure assay (800mg/kg) is relatively high compared to other clinically used CNS compounds. Testing 4-TBB at a lower dose (400mg/kg) did not result in statistically significant effects on seizure (data not shown). To identify a more potent analogue, we exploited Drosophila para bss to screen for anticonvulsive activity of a diverse set of fourteen (synthesised or purchased) 4-TBB analogues (Fig. 4, all structures and chemical properties are shown in sTable 1). We identified 4-(3,5-dimethyl-1H-pyrazol-4-yl)benzoic acid (hereafter termed RAB216) to be potently active against para bss at both 2mM (concentration added to food, Fig. 4A) and at 0.1mM (4-TBB was inactive at this level, data not shown). Analysis of mouse brain exposed to RAB216 (200mg/kg, once per day for 3 days), showed a significant increase in Pum2 expression (p = 0.01), and a smaller (but not significant, p = 0.07) increase in Pum1 (Fig. 5A). We also screened for effect of valproate (VPA, an effective anticonvulsant in the PTZ-induced seizure assay) and saw no change to either Pum1 or 2 (Fig. 5B). Consistent with its heightened potency to increase Pum2 (c.f. Fig. 3C), RAB216 was > 4x more active than 4-TBB in protecting mice against PTZ-induced seizures, being significantly active at 200mg/kg (Fig. 5C-D), a dose which prevented the induction of seizure in 50% of animals tested (p = 0.05, Fig. 5E). By contrast, a repeat of 4-TBB exposure (800mg/kg), in this assay, prevented tonic-clonic seizures in only 30% of animals during the 20min observation period (not significantly different to CTRL, data not shown). This provides confidence that further, yet more active chemical structures, may be discovered. 4-TBB analogues (structures shown in sTable 1) identified a number of active compounds, of which 4-(3,5-dimethyl-1H-pyrazol-4-yl) benzoic acid (RAB216) was the most potent. Relative recovery time (recovery time normalised to para bss run at the same time as drugs) was calculated as a ratio of the treatment group (para bss + compound) recovery time compared to the corresponding untreated group (para bss -compound) from that week of screening. Green denotes a significant reduction in recovery time, orange denotes no change and red a proconvulsant effect. Sodium valproate (VPA, gray bar) was included as an additional control. * = p < 0.05, ** = p < 0 .01, *** = p < 0.001. (unpaired t-tests, between para bss -compound vs. para bss + compound).

Discussion
We report here that 4-TBB and analogues, especially RAB216, are potent anticonvulsants with a novel mode of action that involves up-regulation of the homeostatic regulator Pum. Neuronal homeostasis is likely to assume particular importance in epileptic circuits because of the extreme levels of activity associated with the condition. Notably though, homeostasis has never been specifically targeted for anticonvulsant control. An added appeal for targeting homeostatic mechanisms is that these may, realistically, be expected to impact rather less on normal physiology.
As such, we predict that anticonvulsant strategies, involving neuronal homeostasis, may prove less susceptible to side effects, which are the primary reason for switching anti-epileptic medication in the clinic 23 . This is because homeostatic mechanisms have multiple in-built protective regulatory controls that work to prevent under-or over-activation; in the case of Pum, this protein also targets its own mRNA and at least one of its cofactors 6 .
Whilst multiple forms of neuronal homeostasis have been described, including synaptic scaling and presynaptic regulation of neurotransmitter release 24,25 , the compounds we describe seemingly manipulate firing rate homeostasis, which acts to maintain action potential firing within predetermined and physiologically relevant limits 2 . This is achieved, at least in part, by Pumdependent control of voltage-gated Na + channel synthesis 1 . Analyses of Nav protein levels, in postmortem mouse brains, pre-exposed to 4-TBB, validates this mode-of action, showing reduction in Navs 1.1 and 1.2 but, interestingly, no change in Nav 1.6. Intuitively, one might predict a reduction in Nav1.6 because gain-of function mutations in the encoding gene, Nav 1.6, are associated with hyperactivity and epilepsy 26 . By similar logic, the observed reduction of Nav1.1 is also unexpected given that this channel type predominates in GABAergic inhibitory neurons 27 . Our analysis of these known targets of Pum is, however, relatively crude in treating the whole brain as a single tissue. This approach similarly identifies reduced expression of the GLUR2 AMPA receptor subunit following exposure to 4-TBB. Again, how a reduction in this receptor subunit affects neuronal activity, particularly across the entire brain, is difficult to predict. Glutamatergic synaptic currents, in neurons with reduced expression of GLUR2, exhibit increased deactivation rates which may limit the degree of depolarization induced in the postsynaptic cell 28 . Whilst details remain to be resolved, the changes we observe in Nav and GLUR2 protein levels are consistent with increased Pum expression and, in this regard, serve to strengthen our hypothesis that up-regulation of this homeostatic regulator contributes to the anticonvulsant effect of 4-TBB and RAB216.
We can at present only speculate on how 4-TBB-like molecules mediate an increase in Pum expression. Indeed, in this regard, it is interesting to consider how neurons monitor their activity which, in turn, is transduced to regulate the activity status of intrinsic homeostatic mechanisms. In the case of Pum, we have reported, in Drosophila, that synaptic depolarization regulates expression of p300, a histone acetyltransferase that forms a complex with Mef2. As synaptic depolarization increases, levels of p300 reduce, releasing Mef2 from the complex. Once released, Mef2 binds the dpum promoter and transactivates gene transcription 19 . In mammals, by contrast, the level of Mef2 expression is itself activity-regulated, increasing with depolarization 29 , and analysis of human and mouse pum2 promoters identifies multiple Mef2 binding motifs 19 . p300 is also reported to regulate Mef2 in mammals 30 , but how this protein is influenced by synaptic depolarization has not been described. In mammals, Mef2 also increases the expression of micro-RNAs, including miR-134, which is sufficient to down-regulate expression of Pum2 31,32 . Significantly, block of miR-134, using an antagomir, is anticonvulsive in rodents 33 . Thus, it is conceivable that 4-TBB, and analogues, might act at any level throughout this seemingly complex regulatory mechanism that ensures appropriate expression of Pum proteins. It is expected that levels of Pum are tightly regulated given the requirement to guard against under or over-activity of neuron activity. Indeed, these extensive regulatory and feedback controls, present in Pum-dependent homeostasis, may prove beneficial in exploiting this system for anticonvulsive therapy: minimising potential sideeffects of exposure to 4-TBB or its analogues.
In summary, the study we report here provides a first proof-of-principle that manipulation of neuronal homeostasis, and in particular Pum, provides an exploitable route to suppress seizures and, moreover, may be suitable for the treatment of patients that have drug-refractory seizures.

Acknowledgements
We thank Aoibhinn Kelly, Iona Hayes, Thomas Humphreys and Ceri Hughes who contributed to

Data availability statement
All research data supporting this publication are directly available within this publication.

Materials and correspondence:
Richard.Baines@manchester.ac.uk Seizure behaviour test in Drosophila: Wall-climbing, third-instar larvae (L3), of either sex, were subjected to an electric shock (4V DC, 3s) to induce seizure, with or without previous feeding of compound, as described 34 . Recovery times (RT) are shown which depict the time taken for larvae to recover, evidenced by a full peristaltic wave and normal locomotion. A cut-off time of 420s was used. For compound-feeding studies, eggs were laid on food containing compound (or vehicle, 0.4% DMSO) and larvae were raised (in the presence of drug) until L3. Where experiments were conducted over a number of weeks (e.g. analogue screen shown in Fig. 4), RT was normalized to the para bss (without compound) run each week.
dPum:promoter assay: A dpum promoter-GAL4 line 19 was crossed to attP24 UAS-luciferase flies 35 . Flies carrying the UAS-luciferase transgene alone were used for background controls. Adult flies were allowed to lay eggs in vials containing food with added compound (or vehicle, DMSO) and to develop to L3. Ten L3 CNSs, of either sex, were placed in 100 µl Promega Glo Lysis buffer for each sample, and 5 independent samples collected. CNSs were homogenized, incubated at room temperature (10 min), centrifuged (5 min), and supernatant transferred to a new tube. 30 µl of each sample was then transferred to a well of a white-walled 96-well plate at room temperature, 30 µl Promega Luciferase reagent was added to each well and plates incubated in the dark (10 min).
Luminescence was measured with a GENios plate reader (TECAN, Reading, UK). Values were normalized to total protein concentration, measured using the Bradford protein assay (Bio-Rad, Watford, UK).
Quantitative RT-PCR: QRT-PCR was performed using a SYBR Green I real-time PCR method (Roche, LightCycler® 480 SYBR Green I Master, Mannheim, Germany) as described 36   6Hz seizure-induction: NMRI mice (35−45g, male, Charles River, Chatillon-sur-Chalaronne, France) were used. Prior to the electrical stimulation, 0.5% xylocaine was applied to the cornea to induce local anesthesia and ensure good conductivity. Corneal stimulation (46 mA, 0.2 ms duration pulses at 6 Hz for 3 s) was administered by a constant current device (ECT Unit 57800; Ugo Basile, Comerio, Italy) 40,41 . Acutely evoked 6-Hz seizures were characterized by stun, forelimb clonus, twitching of vibrissae, and/or Straub-tail. For each animal, the total seizure duration was manually recorded. Ip administration of levetiracetam (LEV, 100mg/kg) 1h before seizure induction was used as a positive control 42 . Seizure scoring was carried out live during the experiment. The entire experiment was also video recorded. The researcher was blinded to the experimental groups until full analysis was complete.
Western blot: Whole brain was homogenised in ice-cold buffer (150 mM NaCl, 50 mM Tris-HCl, Chemical synthesis: RAB216 was designed by first screening analogues of 4-TBB and its carboxylic acid derivative RAB102 (data not shown). The structure activity relationship (SAR) revealed that only compounds with a carboxylic acid or aldehyde group directly bonded to the benzene ring were active, even when replaced with isosteric groups. Furthermore, a change of the benzene ring to a pyridine or indole ring, or a change of substituents from para to meta, resulted in complete loss of activity. Electronegative groups were tolerated in the para-position, increasing the likelihood of a compound forming strong intermolecular bonds with its binding partner. Therefore, when designing the second-generation compounds (sTable 1), some analogues included electronegative oxygen and nitrogen atoms. This included RAB216 which contained a pyrazole ring, providing extra interactions to enhance activity. Additionally, analysis of the screen showed that compounds with an element of 3D structure were more active: in RAB216 the two methyl groups attached to the pyrazole group caused it to be twisted by approximately 18°. Reviews of drug libraries suggest that planar compounds are less likely to be biologically active 43 . Compounds, including RAB216, were designed in accordance with Lipinski's rules, such as molecular weight (under 500) and lipophilicity (logP under 5), to give desirable physiochemical properties 44 . All of the compounds screened, including RAB216, fit within these guidelines (sTable 1).
Statistics: Statistical significance was tested using either a Student's two-tailed t-test (paired or unpaired), a one-way or two-way ANOVA followed by post-hoc testing (multiple experimental groups) or Fishers exact test. Level of significance on figures is indicated by * (p ≤ 0.05), ** (p ≤ 0.01), *** (p ≤ 0.001). Figures show means ± SD.