Dose-dependent reversal of KCC2 hypofunction and phenobarbital-resistant neonatal seizures by ANA12

Neonatal seizures have an incidence of 3.5 per 1000 newborns; while hypoxic-ischemic encephalopathy (HIE) accounts for 50–60% of cases, half are resistant to 1st-line anti-seizure drugs such as phenobarbital (PB). Tyrosine receptor kinase B (TrkB) activation following ischemic injury is known to increase neuronal excitability by downregulation of K-Cl co-transporter 2 (KCC2); a neuronal chloride (Cl−) co-transporter. In this study, three graded doses of ANA12, a small-molecule selective TrkB antagonist, were tested in CD1 mice at P7 and P10 following induction of neonatal ischemia by a unilateral carotid ligation. The PB loading dose remained the same in all treatment groups at both ages. Evaluation criteria for the anti-seizure efficacy of ANA12 were: (1) quantitative electroencephalographic (EEG) seizure burden and power, (2) rescue of post-ischemic KCC2 and pKCC2-S940 downregulation and (3) reversal of TrkB pathway activation following ischemia. ANA12 significantly rescued PB resistant seizures in a dose-dependent manner at P7 and improved PB efficacy at P10. Additionally, female pups responded better to lower doses of ANA12 compared to males. ANA12 significantly reversed post-ischemic KCC2 downregulation and TrkB pathway activation at P7 when PB alone was inefficacious. Rescuing KCC2 hypofunction may be critical for preventing emergence of refractory seizures.


EEG power was not a reliable indicator of dose-dependent seizure suppression. Previous studies
have used EEG power to ascertain anti-seizure efficacy of proposed AEDs 9,39 . In this study, EEG power failed to accurately mimic actual EEG seizure burden data ( Power reduction ratios were calculated to normalize each animal's 2 nd hour power to its own 1 st hour baseline (Fig. 3D). All doses of ANA12 + PB significantly reduced EEG power ratios (1 st h to 2 nd h) compared to PB alone [ANA12 0.5 mg/kg: 7.5 ± 1.3 (t 10 = −2.498 p = 0.032), ANA12 2.5 mg/kg: 14.9 ± 3.8 (t 5.578 = −2.932 p = 0.029), ANA12 5 mg/kg: 16.5 ± 4.1 (t 10 = −3.114 p = 0.011); vs PB; independent t-test; Fig. 3D]. EEG power ratio data suggested that all doses of ANA12 were equally effective; however, 0.5 mg/kg had no significant effect on seizure suppression and 5 mg/kg was significantly better than 2.5 mg/kg (Fig. 1B). Therefore, EEG power alone could not detect the dose-dependent efficacy of ANA12 + PB. Repeated measures ANOVA showed that all doses of ANA12 + PB significantly reduced EEG power (Fig. 3E1) at the epoch representing the 15 minute time point after PB injection (F 3,20 = 6.176 P = 0.004, one-way ANOVA; post-hoc, ANA12 0.5 mg/kg: p = 0.031; ANA12 2.5 mg/kg: p = 0.009; ANA12 5 mg/kg: p = 0.010; vs. PB). All other epochs remained non-significant. Percent seizure suppression based on EEG seizure burdens and associated EEG power reduction at this time point (percent seizure suppression vs. EEG power at 75 min within treatment groups) was only significant between ANA12 0.5 mg/kg and 2.5 mg/kg (F 3,40 = 22.83 P < 0.0001, two-way ANOVA; post-hoc, p = 0.0024; Fig. 3E2) indicating erroneous interpretation if EEG power was solely relied upon. Therefore, EEG power failed to demonstrate the differences in seizure burden between doses of ANA12 + PB (Figure supplement 1A-D), and thus failed as an indicator of ANA12's dose-dependent seizure suppression. EEG power alone should not be used to gauge anti-seizure efficacies between animals.
Previously, 3 h EEG data of ANA12 related seizure suppression and the corresponding 3 h post-ligation WB data 38 (Suppl. Figure 6) demonstrated the acute role of ANA12 + PB in rescuing post-ischemic KCC2 degradation. To look at the very early role of ANA12 in the rescue of both PB-refractoriness and post-ischemic KCC2 and pKCC2 degradation in the model before PB was administered (i.e.; 1 h post-ligation) we evaluated WB data at the 1 h time-point (i.e., time point before PB administration). Data showed that ANA12 (5 mg/kg, i.e.; dose with highest anti-seizure efficacy) rescued both KCC2 and pKCC2 at 1 h (Suppl. Figure 2). These data indicate that reversal of refractoriness by ANA12 is dependent on the rescue of the early and immediate prevention of ischemia induced KCC2 hypofunction at P7. ANA12 rescued P10 post-ischemic KCC2 downregulation only at 5 mg/kg. 24 h post-P10 ischemia, KCC2 expression was not significantly different within any treatment group (4E & F). Between groups, PB alone had significantly less contralateral KCC2 expression than naïve controls, therefore failing to rescue KCC2 expression [contralateral KCC2 expression = 73.9 ± 9.6% (F 4,33 = 3.822 P = 0.012, one-way ANOVA; post-hoc, p = 0.036; vs naïve control); Fig. 4E and G]. 5 mg/kg ANA12 significantly rescued KCC2 degradation [% contralateral KCC2: 108.9 ± 7.1% (F 4,33 = 3.822 P = 0.012, one-way ANOVA; post-hoc, ANA12 5 mg/kg: p = 0.018; vs. PB; Fig. 4E and G]. KCC2 downregulation was rescued by all three doses of ANA12 + PB, but there were no significant differences between doses. 24 h post-P10 ischemia, post-ischemic pKCC2 was evaluated ( Fig. 4E-G). Within treatment groups, only 0.5 mg/kg of ANA12 + PB had significantly lower pKCC2 in the right hemisphere compared to the left (t 6 = 2.543 p = 0.044, paired t-test; Fig. 4E,F). Between groups, percent contralateral pKCC2 was not significantly down-regulated in any treatment group, including PB alone, when compared to naïve controls [PB alone: % contralateral pKCC2 = 81.6 ± 11.3%, t 14 = 1.099 p = 0.290, independent t-test; Fig. 4E and G]. There were no significant differences in ratios of pKCC2 to total KCC2 ratios between all treatment groups ( Fig. 4E and H). Neither ischemia nor ANA12 significantly modulated pKCC2/KCC2 ratios, in sharp contrast to P7 ( Fig. 4H; for stats, see Table 3B).
TrkB pathway was not significantly modulated at P10 by ischemia or ANA12. 24 h post-P10 ischemia, the effect of ANA12 + PB on the TrkB pathway was investigated (Fig. 5E-H). At P10, ischemic insult also resulted in a similar upregulation of TrkB, but was only significant in the left hemisphere (TrkB: Left-F 4,32 = 6.782 P ≤ 0.001, one-way ANOVA; post-hoc, p ≤ 0.001; vs naïve control; Fig. 5E,F). No significant upregulation of pTrkB was detected at P10; (Fig. 5F). Only 5 mg/kg of ANA12 + PB significantly rescued ischemia induced upregulation of TrkB in the left hemisphere compared to PB alone, which was not detected for pTrkB (TrkB: Left-F 4,32 = 6.782 P ≤ 0.001, one-way ANOVA; post-hoc, p = 0.004; Fig. 5E,F). Compared to PB alone, ANA12 + PB did not significantly alter percent contralateral TrkB/pTrkB expression (Fig. 5E,G). Neither ischemia nor any dose of ANA12 + PB significantly modulated TrkB or pTrkB in the right ischemic hemisphere at P10. pTrkB/TrkB ratios at P10 following ischemia + PB were modulated differently compared to P7 (i.e.; ratio was lower rather than higher compared to age-matched controls, Fig. 5 compare D vs. H). The higher doses of ANA12 + PB significantly further lowered pTrkB/TrkB ratios at P10 ( Fig. 5H; for stats, see Table 3D) except with 0.5 mg/kg. This indicated that the TrkB-pathway did not play a significant role in post-ischemic seizure generation when PB remained efficacious and no refractoriness was detected.
PLCγ1 phosphorylation not significantly modulated at P10 by ANA12. 24 h post-P10 ischemia, the effect of ANA12 + PB on TrkB associated PLCγ1 expression was investigated ( Fig. 6D-F). PLCγ1and pPLCγ1 expression were not significantly different between all treatment groups. Neither ischemia nor any dose of ANA12 + PB significantly modulated PLCγ1 or pPLCγ1at P10. pPLCγ1/PLCγ1 ratios however were significantly higher in the right ischemic hemisphere at P10, but ANA12 failed to modulate them ( Fig. 6F; for stats, see Table 3F).
At P10, significant post-ischemic vGLUT1 depletion was rescued by ANA12 + PB in a dose-dependent manner Fig. 7E and F; for stats see Table 3). This dose-dependent rescue of vGLUT1 was not evident at P7. At P10, vGAT was significantly depleted in the right hemisphere compared to left in the PB alone group (vGAT: t 6 = 3.306 p = 0.016; paired t-test; Fig. 7E and F). These data indicate vGLUT1 and vGAT were differentially modulated by ischemic insult and ANA12 + PB significantly rescued vGLUT1 depletion at both P7 and P10. At P7 and P10 no significant activation of CREB (pCREB-S133) was detected at 24 h post-ischemia, compared to naïve controls (Figure supplement 3A-D).
KCC2 hypofunction and TrkB activation were age-dependent. Analysis of KCC2 post-ischemic expression indicated unilateral degradation of KCC2.Previous data has shown early degradation of KCC2 (i.e. 3 h) following ischemia in this model 25,38 . To evaluate the dose-dependent effect of ANA12 on this early degradation, the percent contralateral expression of KCC2 to seizure suppression was evaluated. Seizure suppression significantly correlated with the rescue of post-ischemic KCC2 degradation at P7, but not at P10. (r (49) = 0.480, p ≤ 0.001; r (28) = 0.178, p = 0.347 respectively; Spearman's test; Fig. 8A and C). pKCC2 had no significant correlations with seizure suppression at P7 or P10 (pKCC2 vs. seizure suppression: r (38) = 0.154, p = 0.230; r (24) = −0.052, p = 0.802; respectively, Spearman's test; Fig. 8B and D). At both ages, correlations between KCC2 and pKCC2 expression in the left and right hemispheres at 24 h post-ischemia P7 and P10. The ratio of normalized pKCC2 to normalized KCC2 was calculated for the same pup. (Between-group comparisons were done using one-way ANOVA; Post-hoc Bonferroni; *P < 0.05, **P < 0.01, ***P < 0.001 compared to PB-alone. Bars with "#" denote significance between group comparisons to Naïve Control; # P < 0.05, ## P < 0.01, ### P < 0.001). For 1 h KCC2/ pKCC2 WB data see Suppl. were not significant bilaterally. Although pKCC2 is known to play a significant role in KCC2 membrane stability, data showed that when all treatment groups are pooled together, only total KCC2 was significantly associated with seizure suppression outcomes in this model at P7 but not at P10.
To further investigate the sex-dependent efficacy of ANA12 at P7, we analyzed the protein data by sex. At P7, ischemia significantly upregulated TrkB in the ischemic hemisphere in males but not females (Males: TrkB: Right, F 4,29 = 6.693 P = 0.007, one-way ANOVA; post-hoc, p = 0.020; vs. naïve control group). TrkB activation has been linked to KCC2 downregulation 26,42 . These data may explain the sex-differential activity of ANA12 + PB; however, further studies are needed. At P10, the age when PB was effective, there were no significant sex differences in KCC2 or TrkB expression.

Discussion
This study has several key findings: 1. Small-molecule TrkB antagonist, ANA12, reversed post-ischemic PB-resistant seizures in a dose-dependent manner; 2. Rescue of PB-resistance was associated with a dose-dependent rescue of KCC2 downregulation; 3. The pKCC2-S940 site played a significant role in the rescue of KCC2 membrane stability associated with ANA12 + PB anti-seizure efficacy; 4. The TrkB pathway was significantly activated following neonatal ischemia at P7 with upregulation of both pTrkB and its downstream activation site pPLCγ1, each were rescued to control levels; 5. Reversal of ischemia induced pTrkB activation correlated significantly with percent seizure suppression at P7, 6.TrkB pathway activation at P10 via pTrkB-T816 was not significant, nor was there PB-resistance at that age. 7. Ischemia induced vGLUT1 depletion following insult which was significantly rescued by ANA12 + PB, in contrast, vGAT showed no significant modulation.
Ischemic insults leads to rapid increase in BDNF, and BDNF binding to the TrkB receptor is responsible for the pathophysiological downregulation of KCC2 [26][27][28]42 . K252a, a TrkB receptor antagonist, was shown to produce a significant increase in KCC2 levels in organotypic hippocampal slice cultures 26 . Another well studied area is the intrinsic functional modulation of KCC2 by phosphorylation at various residues 23,45 . The dephosphorylation of the pKCC2-S940 site triggers the internalization of membrane bound KCC2, thus reducing surface expression and hence function 23,46 . The acute, specific and transient post-ischemic inhibition of the TrkB receptor with ANA12 can help prevent acute KCC2 downregulation 25,38 . However, the age-, sex-and dose-dependence underlying the mechanism for TrkB mediated KCC2 downregulation in neonates are unknown.
KCC2, in addition to its role in maintaining the Cl-transmembrane gradient, has also been shown to play an independent role in cell-survival, synaptogenesis, and AMPA receptor insertion at excitatory synapses [47][48][49] . KCC2 modulation may be a beneficial pharmacological target for excitotoxic insults during development that result in increased susceptibilities to long-term sequelae like epilepsy 50 , schizophrenia 51 , and disorders of dendritic spine formation 52 .
KCC2 is highly expressed at excitatory synapses, and is essential for glutamatergic synapse development and function. The suppression of KCC2 activity reduces the localization of GluR1 composed AMPA receptors at the synapse, and subsequently reduces the amplitude of EPSCs 53 . Increased glutamatergic signaling also increases KCC2 membrane dynamics at excitatory synapses, as NMDA activation leads to Ca 2+ dependent dephosphorylation of KCC2-S940 54 . Therefore, in addition to being the chief neuronal chloride extruder, KCC2 also curbs hyperexcitability by attenuating the response to glutamatergic signaling at excitatory synapses.
In contrast, hyperpolarizing GABA A signaling stabilizes KCC2 at inhibitory synapses 55 . Other pathways known to modulate KCC2 function include WNK1, a chloride sensitive kinase, that is activated by low intracellular Cl − levels, and phosphorylates KCC2 at T906 and T1007 55 . Wright et al. 2017 have demonstrated that metabotropic GABA B receptor activation lead to clathrin-mediated endocytosis of KCC2 56 . The calpain mediated pathway activated by prenatal hypoxia-ischemia can also degrade KCC2, erythropoietin has demonstrated an ability to rescue this degradation 57 . Therefore, KCC2 hypofunction in neonatal ischemia may alter intrinsic excitability of neurons at both excitatory and inhibitory synapses. Further studies are required to determine KCC2's role within specific cell types, and their associated injury from hypoxic-ischemic insults.
Glutamatergic neurotransmission is highly dependent upon the function of vGLUTs to transport glutamate into synaptic vesicles. One of the features of VGLUT function is its Cl − dependence, which is not fully understood. Cl − modulates vGLUT1, and thus modulates glutamate transport into vesicles 60 . The allosteric activation of vGLUT1 and vGLUT2 is inhibited by ketone bodies, consistent with clinical outcomes reported for the ketogenic diet used to treat children with drug-resistant epilepsies 61,62 . This suggests a possible role of intracellular chloride concentrations as a second messenger in the presynaptic terminal, modulating vGLUT1.Further studies are required to understand the in vivo intracellular Cl − concentrations at the presynaptic terminal, how vGLUT1 modulation affects systemic neurotransmission, and the direct effects of seizures on both presynaptic intracellular chloride levels and vGLUT1. In this study, ischemia induced seizures depleted vGLUT1 but had no effect on vGAT, consistent with previous reports 35 . ANA12 + PB was able to rescue this depletion of VGLUT1 at all of the doses tested, supporting that vGLUT depletion is mediated by seizures.
This study provides novel insights into the unknown developmental influence on the TrkB pathway. TrkB pathway modulation played a significant role in the dose-dependent anti-seizure efficacy of ANA12 for PB-resistant seizures at P7, but not for the PB-responsive seizures at P10. In contrast, ANA12 played a significant role in the (F) PLCγ1 to pPLCγ1 ratio at P10. Significance to PB: *P < 0.05, **P < 0.01, ***P < 0.001. Significance to naïve control: # P < 0.05, ## P < 0.01, ### P < 0.001. Ipsilateral vs. contralateral: γ P < 0.05, γγ P < 0.01, γγγ P < 0.001. rescue of post-ischemic dephosphorylation of pKCC2-S940 at both P7 and P10. The age-dependent efficacy of PB for neonatal ischemic seizures may underlie the developmental switch from GABA acting as a depolarizing agent in immature brains, to its hyperpolarizing function in mature brains 14 in this CD-1 mouse model 25 . ANA12 is the first novel low-molecular weight TrkB antagonist 36 that crosses the blood brain barrier. It binds to the extracellular domain of TrkB and prevents BDNF-induced TrkB activation, but does not prevent the biological action of BDNF on NGF or NT-3 on TrkA-and TrkC-expressing cells. ANA12 has been shown to inhibit TrkB in the brain after systemic administration without any significant adverse effects, even with multiple dosing protocols in adult mice 36 .
This study demonstrates that power is not a reliable proxy for seizures, especially when recording from free moving subjects such as seizing pups. In this study, electrode sites differ slightly in placement and therefore no between subject power analysis would be meaningful. Additionally, repetitive seizures change in amplitude and frequency as a function of time; this is especially true for ischemic seizures 4,7,63 . Therefore, EEG power has an abbreviated role in acute seizure analysis (i.e. high power or lower power seizure) and is only meaningful within a subject, and not between subjects. Our findings are similar to reports demonstrating low sensitivity to seizure detection algorithms that rely on power, yet high accuracy on algorithms that utilize time and frequency characteristics 64 .
The TrkB pathway plays a critical role in the post-ischemic emergence of refractory seizures. This characterized CD1 mouse model of PB-resistant seizures was shown to be age-dependent 25 and this study showed that the TrkB pathway modulation was also significantly age-dependent. KCC2 downregulation occurred only in the ischemic hemisphere indicating a direct effect of ischemia; in contrast pKCC2 downregulation occurred bilaterally indicating that this may occur due to repetitive seizures. The pKCC2-S940 phosphorylation site also seemed to play a role at P7 and not at P10, indicating that its role in KCC2 hypofunction was age-dependent which may be associated with the seizure severity and PB-refractoriness that occurs at P7. It is not clear whether the rescue of pKCC2-S940 with ANA12 is a direct effect of ANA12 interacting with TrkB or an indirect effect by preventing TrkB mediated KCC2 downregulation. However, KCC2 S940A mutants show functional deficits following transient exposure to glutamate, suggesting that the S940 site plays an important role in KCC2 activity during excitotoxic conditions, in contrast to basal conditions 45 . Future studies with TrkB mutant mice may help answer this question. However, the data presented here demonstrate that ANA12, a small-molecule TrkB antagonist rescues Quantification of blots shown in C. vGLUT1 is downregulated by ischemia and rescued in a dosedependent manner by ANA12 + PB. vGAT is unaffected by ischemia. Significance to PB alone: *P < 0.05, **P < 0.01, ***P < 0.001. Significance to naïve control: # P < 0.05, ## P < 0.01, ### P < 0.001. Significance to 5 mg/kg of ANA12 + PB: @ P < 0.05, @@ P < 0.01, @@@ P < 0.001. Ipsilateral to contralateral: γ P < 0.05, γγ P < 0.01, PB-refractoriness and KCC2 hypofunction by reliably and significantly reversing ischemia induced TrkB-pathway activation at P7 in a dose-dependent manner.

Materials and Methods
All experimental procedures were conducted in compliance with guidelines by the Committee on the Ethics of Animal Experiments, Johns Hopkins University (Permit Number: A3272-01) and all protocols were approved by the Animal Care and Use of Committee (IACUC) of Johns Hopkins. All litters of CD1 mice with dams were purchased from Charles River Laboratories Inc. (Wilmington, MA.). Newly born litters of pups (n = 10) with dams were delivered at postnatal days old (P3 or P4) and allowed to acclimate. Food and water were provided ad libitum. Equal numbers of male and female pups were used in the study. Sample sizes are included in Table 1.
Surgical procedure for ischemic insult and sub-dermal EEG electrode implantation. The surgical protocol was similar to the previously published work 25,65 . At P7 or P10, animals were subjected to permanent unilateral ligation (without transection) of the right common carotid artery using 6-0 surgisilk (Fine Science Tools, BC Canada) under isoflurane anesthesia. The outer skin was closed with 6-0 monofilament nylon (Covidien, MA), and lidocaine was applied as local anesthetic. Under continued anesthesia, animals were then implanted with 3 sub-dermal EEG scalp electrodes: 1 recording and 1 reference overlying the bilateral parietal cortices, and 1 ground electrode overlying the rostrum. Wire electrodes made for use in humans (IVES EEG; Model # SWE-L25 -MA, IVES EEG solutions, USA) were implanted sub-dermally and fixed in position with cyanoacrylate adhesive (KrazyGlue). Pups were then allowed to recover from anesthesia over a few minutes. Pups were then tethered to a preamplifier by connecting to the sub-dermal electrodes within a recording chamber for 2 h of continuous video-EEG recording, maintained at 36 °C with isothermal pads. At the end of the recording session, sub-dermal electrodes were removed, and the pups were returned to the dam. The average duration of anesthesia for both ligation and electrode implantation surgery was 16.18 ± 4.37 min. There is a known mortality rate of ~10-20% associated with the surgical procedure of carotid-ligation and severe seizures in the model 66 . The mortality rates for the pups 24 h after surgery were n/n = 7/149 pups at P7 (5 males and 2 females) and n/n = 0/49 pups at P10, and were not significantly different, by age nor by sex (p = 0.20 and p = 0.70 respectively; Fisher's exact test, two-tailed). Mortality rates following the surgical procedure were also not significantly different by treatment (ligated control vs. treated group; p = 0.19; Fisher's exact test, two-tailed).  Fig. 1A. ANA12 was dissolved in 5% dimethyl sulfoxide (DMSO) and stored at −20 °C in aliquots. ANA12 dose-range was chosen based on the efficacious dose previously published in our pilot 'proof of principle" paper 25 and expanded based on safety studies that determined non-toxic higher doses in mice with chronic dosing protocols 36 . A previous study showed that DMSO alone did not alter seizure burdens in this model, hence a vehicle alone treatment group was not included here 38 . PB (25 mg/kg; Sigma-Aldrich; Cat. No. P5178) was dissolved in phosphate buffered saline and injected 1 h after the ANA12 injections. The 25 mg/kg PB loading dose replicated a pilot study protocol and other similar pre-clinical studies 9,25 . Recent clinical trials have also shown that when a loading dose of PB at 20 mg/kg was ineffective, a second follow-on dose of 20 mg/kg over a 12 h period 4,67 was ineffective in curbing neonatal seizures therefore no additional doses of PB were administered here.

In vivo synchronous video-EEG recording and analyses. EEG recording was acquired using Sirenia
Acquisition software (v 1.6.4) with synchronous video capture (Pinnacle Technology Inc. KS, USA). Data acquisition was done with sampling rates of 400 Hz that had a pre-amplifier gain of 100 and a 0.5 Hz to 50 Hz band pass filter. Data were scored by binning EEG in 10 sec epochs. Similar to our previous study 25 , seizures were defined as electrographic ictal events that consisted of rhythmic spikes of high amplitude, diffuse peak frequency of ≥7-8 Hz (i.e.; peak frequency detected by automated spectral power analysis) lasting ≥6 seconds (i.e.; longer than half of each 10 sec epoch on the manual scoring screen within the module). Short duration burst activity lasting <6 seconds (brief runs of epileptiform discharges) were not included for seizure burden calculations similar to previous studies in the model 25 (35)(36)(37)(38)(39)(40)(41)(42)(43)(44)(45)(46)(47)(48)(49)(50). Total EEG power was calculated for the first hour (3600 seconds) and second hour of recording, post-PB injection (3600 seconds). The total EEG power for every 5 minutes (24 epochs of 300 seconds) was calculated for each 2 hour recording. From each treatment group, 6 randomly chosen P7 pups (after excluding pups with electrical noise related EEG artifacts) were used for EEG power analysis [PB