Zinc regulates a key transcriptional pathway for epileptogenesis via metal-regulatory transcription factor 1

Temporal lobe epilepsy (TLE) is the most common focal seizure disorder in adults. In many patients, transient brain insults, including status epilepticus (SE), are followed by a latent period of epileptogenesis, preceding the emergence of clinical seizures. In experimental animals, transcriptional upregulation of CaV3.2 T-type Ca2+-channels, resulting in an increased propensity for burst discharges of hippocampal neurons, is an important trigger for epileptogenesis. Here we provide evidence that the metal-regulatory transcription factor 1 (MTF1) mediates the increase of CaV3.2 mRNA and intrinsic excitability consequent to a rise in intracellular Zn2+ that is associated with SE. Adeno-associated viral (rAAV) transfer of MTF1 into murine hippocampi leads to increased CaV3.2 mRNA. Conversely, rAAV-mediated expression of a dominant-negative MTF1 abolishes SE-induced CaV3.2 mRNA upregulation and attenuates epileptogenesis. Finally, data from resected human hippocampi surgically treated for pharmacoresistant TLE support the Zn2+-MTF1-CaV3.2 cascade, thus providing new vistas for preventing and treating TLE.

E pilepsy affects 1% of individuals of all ages, socioeconomic backgrounds and races, and is the second most common cause of mental disability, particularly among young adults, accounting for a worldwide disease burden similar to that of breast cancer in women and lung cancer in men [1][2][3] . Chronic recurrent seizures often originate in the temporal lobe (temporal lobe epilepsy (TLE)) and are pharmacoresistant in approximately a third of the patients. Surgical removal of the epileptic focus, albeit highly effective, represents a therapy option for only a fraction of patients 4 . Often, TLE develops as a consequence of a brain disease or of an acute brain insult (acquired or 'symptomatic' TLE) via a multifaceted process referred to as epileptogenesis 5 . Intriguingly, a single episode of status epilepticus (SE) can induce the structural and functional alterations that lead to the emergence of chronic recurrent seizures. Identifying the key epileptogenic mechanisms is essential for devising new treatments to prevent or attenuate the development of TLE.
Recent experimental and clinical evidence suggests that acquired 'transcriptional channelopathies' play a role in epileptogenesis, as well as in the pathogenesis of other neurological disorders [6][7][8][9] . Thus, in rodents, SE induced chemically with pilocarpine (pilocarpine-SE) causes a marked increase in the propensity for intrinsic bursting in hippocampal CA1 pyramidal cells, particularly in the early phase of epileptogenesis 10 . Pharmacological analyses with subtype-selective blockers of voltage-dependent Ca 2 þ channels (VDCCs) disclosed the involvement of a Ni 2 þ -sensitive-T-type Ca 2 þ current (I CaT ) in this aberrant activity 11,12 . Congruently, I CaT was threefold upregulated early after SE, whereas other Ca 2 þ currents were unchanged or even reduced 12 . Furthermore, a significant upregulation of Ca V 3.2 mRNA, but not that of other T-type Ca 2 þ -channel a 1 mRNAs, was present in SE-experienced neurons and translated to an increase in Ca V 3.2 protein level 13 . Interestingly, the emergence of epileptic seizures was strongly attenuated in Ca V 3.2 knockout mice subjected to pilocarpine-SE 13 , indicating that transcriptional Ca V 3.2 upregulation may play a pivotal role in epileptogenesis.
It has been shown that SE causes a rise in intracellular free Zn 2 þ concentrations ([Zn 2 þ ] i ) in pyramidal cells 14 and we have recently demonstrated that Zn 2 þ , although acutely and reversibly blocking T-type channels 15 , induces a long-term upregulation of I CaT in hippocampal pyramidal cells in vivo 16 . Therefore, we have aimed at delineating the signalling cascade linking the increase in [Zn 2 þ ] i to Ca V 3.2 promoter activation. Our results in vitro indicate that this link is mediated by metalregulatory transcription factor 1 (MTF1). Supporting data obtained from 'viral transgenic' mice and human TLE hippocampi suggest that MTF1 may be a target for treatment strategies aimed at impeding the development of epilepsy following acute brain insults.

Results
Selective induction of Ca V 3.2 expression by Zn 2 þ . To unravel the signalling cascades underlying Zn 2 þ -induced I CaT upregulation, we first examined the effect of elevating intracellular Zn 2 þ ([Zn 2 þ ] i ) on VDCC gene expression in neural NG108-15 cells. A large increase in [Zn 2 þ ] i was triggered by incubating the cells with a solution containing 200 mM Zn 2 þ (a concentration that can be reached in pathological conditions such as ischemia, seizures and brain trauma 17 ) and high K þ (50 mM to depolarize the cells 18,19 ). Only this combination (further referred to as K þ þ Zn 2 þ ), but not with K þ or Zn 2 þ solutions alone, caused an increase in [Zn 2 þ ] i (Fig. 1a,b). None of these treatments affected cell viability (Fig. 1c).
Next, we analysed the mRNA expression levels of the neuronal L-type (Ca V 1.2 and Ca V 1.3), P/Q-type (Ca V 2.1), N-type (Ca V 2.2), R-type (Ca V 2.3) and T-type (Ca V 3.1, Ca V 3.2 and Ca V 3.3) VDCCs in NG108-15 cells incubated with basal, K þ or K þ þ Zn 2 þ solutions. In basal condition, high expression levels were obtained for Ca V 3.2 and to a lesser degree for Ca V 2.2 and Ca V 3.3 (Fig. 1d). Incubation with K þ þ Zn 2 þ solution resulted in a significant upregulation of Ca V 3.2 mRNA levels, but did not affect mRNA levels of other VDCCs that are sensitive to block by submolar concentrations of Ni 2 þ (Fig. 1e, Supplementary Fig. 1). Furthermore, incubating NG108-15 cells with K þ solutions containing other divalent cations (0.5 mM Ni 2 þ or 1 mM Cu 2 þ ) did not affect Ca V 3.2 mRNA expression ( Supplementary Fig. 2). Thus, a rise in [Zn 2 þ ] i uniquely and selectively enhances Ca V 3.2 mRNA expression, an effect that is expected to cause an increase in Ca V 3.2 protein level. Indeed, immunoblotting detected the Ca V 3.2 protein under basal conditions and revealed a significant increase of Ca V 3.2 protein levels after exposure of the cells to K þ þ Zn 2 þ solution (Fig. 1f,g).
Increase of I CaT by Zn 2 þ . We next used whole-cell patch-clamp recordings of Ca 2 þ currents in NG108-15 cells to investigate whether the Zn 2 þ -induced increase in Ca V 3.2 protein is reflected functionally in a larger I CaT . NG108-15 cells were incubated in either K þ or K þ þ Zn 2 þ solutions and Ca 2 þ currents were recorded. Representative examples of I CaT in a control cell and in a cell exposed to K þ þ Zn 2 þ solution are shown in Fig. 2a. We found that exposing the cells to K þ þ Zn 2 þ solution, but not to K þ solution led to a significant increase in I CaT (Fig. 2b,c). This increase was not accompanied by a change in voltage dependence of activation or inactivation (Fig. 2d, Supplementary Table 1). Expectedly, application of 100 mM Ni 2 þ , a dose potently inhibiting Ca V 3.2 channels (grey traces in Fig. 2e), almost completely reduced I CaT in both basal and K þ þ Zn 2 þ conditions (Fig. 2f). Altogether, these findings are congruent with the notion that Zn 2 þ -induced increase in I CaT reflects an increase in functional Ca V 3.2 channels. Zn 2 þ -induced activation of the Ca V 3.2 promoter. We next sought to identify the molecular mechanisms underlying the Zn 2 þ -induced upregulation of Ca V 3.2. To that end, we made use of the previously identified Ca V 3.2 promoter 20 , the activity of which strongly correlates with endogenous Ca V 3.2 mRNA levels. We found that treating the cells with K þ þ Zn 2 þ solution, but not with K þ solution, resulted in a significant Ca V 3.2 promoter activation (Fig. 3a). This effect was enhanced by raising Zn 2 þ concentration above 200 mM. We further examined whether Zn 2 þ -induced activation of the Ca V 3.2 promoter reverses on Zn 2 þ removal. To that end, we exposed cells having been incubated in a K þ þ Zn 2 þ solution to N,N,N 0 ,N 0 -tetrakis (2-pyridylmethyl)ethane-1,2-diamine (TPEN), a cell permeable Zn 2 þ chelator. Intriguingly, TPEN significantly reduced Ca V 3.2 promoter activity. In contrast, basal activity of the Ca V 3.2 promoter was unaffected by TPEN (Fig. 3b). These data suggest that enhanced activation of the Ca V 3.2 promoter requires a sustained increase in [Zn 2 þ ] i . We next delineated the region responsible for Zn 2 þ -induced Ca V 3.2 promoter activation by analysing Ca V 3.2 promoter deletion luciferase reporter constructs (Ca V 3.2-1020, -312 and -105) (ref. 20). Increases in [Zn 2 þ ] i resulted in a significant Ca V 3.2 promoter activity only for the longest deletion fragment (Fig. 3c) suggesting that the Zn 2 þ -responsive regulatory element is located in the genomic between 312 and 1,020 upstream of the Ca V 3.2 ATG. These findings indicate that Ca V 3.2 gene regulation is under control of a Zn 2 þ -responsive transcription factor 21 . The only currently known transcription factor fulfilling these criteria in mammals is MTF1. MTF1 increases Ca V 3.2 promoter activation and I CaT . Subsequent bioinfomatic analyses revealed a surprising accumulation of potential binding sites (metal-responsive elements (MREs)) for MTF1 (Table 1) 22 . One of the bioinformatically detected MREs resided within the above identified Zn 2 þ -inducible minimal Ca V 3.2 promoter region (Fig. 4a), supporting a role for MTF1 in the Zn 2 þ -dependent Ca V 3.2 mRNA expression. To test if MTF1 indeed is able to stimulate the Ca V 3.2 promoter, we analysed Ca V 3.2 promoter activity after MTF1 overexpression in NG108-15 cells and primary hippocampal neurons. Indeed, MTF1 overexpression increased Ca V 3.2 promoter activity in both cell types to similar levels as observed after stimulation with K þ þ Zn 2 þ (Fig. 4b). Whole-cell patch-clamp recordings revealed that overexpression of MTF1 in NG108-15 cells led to a substantial increase of I CaT (Fig. 4c). These augmented currents were almost completely blocked by 100 mM Ni 2 þ (Fig. 4d), indicating that they are generated by Ca v 3.2 subunits.
MTF1 binds a Zn 2 þ -sensitive MRE in the Ca V 3.2 promoter. We then probed whether the Zn 2 þ -inducible genomic fragment of the Ca v 3.2 promoter (Fig. 4a) corresponds to the genomic region of the Ca v 3.2 promoter with maximal MTF1 responsiveness. We therefore analysed the activity of three Ca V 3.2 promoter deletion fragments in the presence of MTF1. Only the Ca V 3.2 promoter fragment harbouring the MRE located in the Zn 2 þinducible region was activated by MTF1 overexpression (Fig. 5a). Subsequent mutation of this MRE significantly reduced the Ca V 3.2 promoter activity after MTF1 overexpression (Fig. 5b), indicating that this MRE is indeed responsible for the MTF1-induced Ca V 3.2 upregulation. In addition, chromatin immunoprecipitation (ChIP) analysis of NG108-15 cells, as well as of mice hippocampi, revealed binding of MTF1 to the Zn 2 þsensitive MRE in the Ca V 3.2 promoter region (Fig. 5c).
To prove unequivocally that the Zn 2 þ -induced Ca V 3.2 promoter activation is mediated by MTF1, a dominant-negative form of MTF1 (MTF1DC) 23 was co-transfected with the Ca V 3.2-1188 reporter plasmid. MTF1DC is still able to bind MREs but does not possess the ability to activate transcription, thereby blocking MREs from wild-type MTF1. Treatment with K þ þ Zn 2 þ solution of cells overexpressing MTF1DC resulted in a complete repression of Zn 2 þ -induced Ca V 3.2 promoter activation (Fig. 5d). Therefore, MTF1 appears necessary and sufficient to mediate the stimulatory effects of Zn 2 þ on the Ca V 3.2 promoter.
MTF1 upregulates hippocampal Ca V 3.2 mRNA in vivo. It was previously shown that pilocarpine-SE induces an increase in [Zn 2 þ ] i in the somata of CA1 pyramidal cells 14 , likely due to the release of intracellularly bound Zn 2 þ (ref. 24). We investigated in rats the time course of this increase using TFL staining of hippocampal slices resected from control and SE-experienced animals ( Fig. 6a; CA1, CA3, hilus and granular layer of the dentate gyrus). The strongest increase in TFL staining was observed in the CA1 region (Fig. 6b). TFL-positive pyramidal cells first appeared at 1 day after SE, their incidence peaking 2-4 days after SE and declining thereafter (Fig. 6b,c). These data suggest an association between the SE-induced rise in [Zn 2 þ ] i and the SE-induced increases in Ca V 3.2 mRNA and protein, as well as in I CaT (ref. 13  group) or rAAV-CMV-Venus particles (control group) into area CA1 of mice hippocampi. The mRNA isolated from hippocampi of both groups revealed significantly increased MTF1 mRNA expression in the former group, indicating efficient viral transduction. Correspondingly, we observed Ca V 3.2 mRNA expression to be significantly increased in the MTF1 group versus the control group (Fig. 6d). In addition, in vivo imaging using infrared fluorescent proteins (iRFP) under control of the Ca V 3.2 promoter (Fig. 6e,f) also showed the activation of the Ca V 3.2 promoter after transduction with MTF1 and a remarkably similar activation after intrahippocampal injection of Zn 2 þ (Fig. 6g-j). These data thus show that MTF1 and Zn 2 þ activate the Ca V 3.2 promoter also in vivo. We next similarly injected mice with MTF1DC. We injected either rAAV-CMV-MTF1DC-IRES-Venus (MTF1DC group) or rAAV-CMV-Venus particles (control group) into area CA1 of mice hippocampi (Fig. 7a). Two weeks after injection, mice were either subjected to pilocarpine-SE or sham treated. The mice were killed 3 days thereafter, that is, at the time point of maximal Ca V 3.2 mRNA increase after SE 13 . Interestingly, overexpression of MTF1DC significantly reduced the SE-induced increase in Ca V 3.2 mRNA peak during early epileptogenesis (Fig. 7b), further indicating that MTF1 mediates between the SE-induced rise in [Zn 2 þ ] i and the transcriptional Ca V 3.2 upregulation.
Given its key role in SE-induced Ca V 3.2 transcription, we also studied whether pilocarpine-SE affects MTF1 expression levels. We found that the levels of MTF1 mRNA significantly increased 6 and 12 h after SE, and returned to basal values thereafter (Fig. 7c).
Interfering with MTF1 attenuates seizure development. We have previously shown that mice lacking Ca V 3.2 manifested a much milder form of chronic TLE 13 . We therefore expected that interfering with the Zn 2 þ -MTF1-Ca V 3.2 cascade would also exert an antiepileptogenic action. We tested this prediction in mice transduced with MTF1DC as described above. With respect to the acute SE induced by pilocarpine, we found no differences in electroencephalography (EEG) recordings between MTF1DC mice and control animals before, during and after the SEs (Fig. 8a). Likewise, the two groups of mice were similar with respect to the latencies to the first acute seizure and to the onset of SE (Fig. 8b). These results show that the two groups of mice likely experience SEs of identical intensities. Further EEG monitoring indicated that spontaneous seizures emerged in both groups of mice (representative examples of EEGs shown in Fig. 8c). However, seizure frequency was substantially lower in MTF1DC mice compared with control animals (Fig. 8d,e). Intriguingly, the extent of neurodegeneration found in MTF1DC mice was similar to that found in control animals, despite the lesser seizure frequency in the former group ( Supplementary Fig. 3).
Increased MTF1 and Ca V 3.2 expression correlate in HS. To assess the potential role of the Zn 2 þ -MTF1-Ca V 3.2 cascade in  human TLE, we analysed hippocampal MTF1 and Ca V 3.2 expression levels in pharmacoresistant TLE patients with hippocampal sclerosis (HS) versus patients with 'lesion-associated' TLE (Supplementary Note 1). We found a strong positive correlation between MTF1 and Ca V 3.2 expression levels in both groups of patients (Fig. 9b, Supplementary Fig. 4). These data indicate that the correlation of expression between MTF1 and Ca V 3.2 is a rather stable phenomenon, especially when considering patients heterogeneity with respect to endophenotypes (for example, hippocampal damage, time point after seizure onset, etc.) and genetic background. Intriguingly, both MTF1 and Ca V 3.2 mRNA expression levels were substantially higher in TLE patients with HS compared with those with 'lesion-associated' TLE (Fig. 9c). This difference may reflect differences in seizure frequencies or intensities between the two groups or in other factors regulating [Zn 2 þ ] i .

Discussion
Here we describe a novel mechanism of neuronal plasticity, which we refer to as the  different components of the Zn 2 þ -MTF1-Ca V 3.2 cascade that lead to Ca V 3.2 upregulation. We found that [Zn 2 þ ] i increases, acting via MTF1, upregulate Ca V 3.2 promoter activity but do not affect transcription of other T-type Ca 2 þ channel subunits. This intriguing selectivity is due to the fact that only Ca V 3.2, but not Ca V 3.1 and Ca V 3.3, contain MREs in the 1.5-kb genomic region upstream of the start ATG. Indeed, gene promoters that harbour MREs are sparsely distributed throughout mammalian genomes 25 . The only ones known to be regulated by MTF1 in a [Zn 2 þ ] i -dependent manner are the genes encoding for metallothioneins. These are small, cysteine-rich proteins with a high affinity for Zn 2 þ and other heavy metals 26 . By increasing the expression of metallothioneins, MTF1 acts in a negative feedback manner to facilitate removal of excess free Zn 2 þ (ref. 27).
In the pilocarpine-SE model, deleting Ca V 3.2 not only reduced the frequency of recurrent seizures in the chronic stage but also strongly protected the hippocampus from SE-induced neurodegeneration 13 . Here we show that transducing hippocampi with MTF1DC also reduced chronic seizures frequency, but did not ARTICLE prevent neurodegeneration. This discrepancy may be due to the fact that MTF1DC overexpression also interferes with MTF1mediated upregulation of metallothioneins, thereby allowing [Zn 2 þ ] i to increase to toxic levels. To overcome this side effect, and to unequivocally prove that the reduction of seizures after treatment with MTF1DC is due to a direct effect on the Ca V 3.2 promoter, the Zn 2 þ -sensitive MRE in the Ca V 3.2 promoter could be genetically modified in vivo using the CRISPR/Cas complex 28 . Subsequent analysis of these mice in the pilocarpine-SE model will then reveal whether they experience less seizure activity as well as a reduced neurodegeneration. Our findings suggest that pilocarpine-SE evokes the Zn 2 þ -MTF1-Ca V 3.2 cascade by inducing a rise in [Zn 2 þ ] i . However, the mechanism coupling SE to [Zn 2 þ ] i increase is yet unknown. During the intense neuronal activity underlying SE, labile Zn 2 þ is released from glutamatergic terminals and may enter postsynaptic neurons via multiple routes 14 . Alternatively, Zn 2 þ tightly bound to metallothioneins may be released by action of nitric oxide 29,30 , whose production is markedly increased during SE 31 . In either case, another potential strategy to impede the Zn 2 þ -MTF1-Ca V 3.2 cascade would be the early application of cell permeable Zn 2 þ chelators. However, given that Zn 2 þ is mandatory for many critical cell processes 32 , its chelation might lead to deleterious effects 33 . An alternative strategy to impede the Zn 2 þ -MTF1-Ca V 3.2 cascade would be to interfere with nitric oxide accumulation by application of nitric oxide synthase inhibitors or scavengers. Pharmacological targeting of Ca V 3.2 blockers early after pilocarpine-SE may also prove to be antiepileptogenic, and selective Ca V 3.2 blockers are becoming available 34 . Alternatively, MTF1 may be targeted pharmacologically. The fact that activation and promoter binding of MTF1 require phosphorylation 23 development of inhibitory drugs, for example, by small molecule library screening 36 . As perspective, we suggest that pharmacological interventions targeting the Zn 2 þ -MTF1-Ca V 3.2 cascade may prove as intriguing future option for treating pharmacoresistant TLE.

Methods
Bioinformatic analysis and plasmids. MREs were identified using the software tool PoSSuMsearch 37 with position-specific scoring matrices from the TRANSFAC database 38 . The mammalian expression vectors pCDNA3-HA-mMTF1 and pCDNA3-MTF1-EcoRl (dominant negative; MTF1DC) were kindly provided by Carl Seguin (Québec) and Guy J. Rosnan (Fred Hutchinson Cancer Center, Seattle).
The Ca V 3.2-1020-MRE-mut reporter plasmid was made by mutating the Zn 2 þsensitive MRE in the Ca V 3.2-1020 luciferase promoter fragment 20 . For this, the first three nucleotides of the MRE consensus sequence (TGC; Table 1)  The pAAV-Ca V 3.2-luciferase construct has been described previously 39 . For the AAV-RL-TK control plasmid, the RL-TK cassette (Promega, Mannheim, Germany) was amplified with NotI overhang and cloned in NotI-digested pAAV-MCS. The pAAV-Ca V 3.2-iRFP 713 was made by exchanging the luciferase from pAAV-Ca V 3.2-venus with iRFP 713 (Addgene clone #31857) using HindIII and BglII restriction sites. All AAV-cloning procedures were performed in Stbl2 bacteria (Life Technologies, Germany) to minimize recombination events. Plasmid sequences were verified by sequencing analysis. Integrity of the inverted terminal repeats was confirmed by SmaI restriction analysis.
Cell cultures. Several types of cultured cells were used in this study. NG108-15 cells (American Type Culture Collection HB-12317) were maintained at 37°C and 5% CO 2 in DMEM supplemented with 10% (v/v) heat-inactivated FCS (Invitrogen) 100 units per ml penicillin/streptomycin, 2 mM glutamine and 1 Â HAT (sodium hypoxanthine, aminopterin and thymidine; Invitrogen). If not stated otherwise, NG108-15 cells were seeded in 24-well plates with 60,000 cells per well. HEK293 cells stably transfected with human Ca V 3.2 (kindly provided by Ed Perez-Reyes, University of Virginia, Charlottesville, VA, USA) and HEK293-AAV cells (#240073, Stratagene, La Jolla, CA) were kept in high-glucose DMEM supplemented with 10% FCS (Invitrogen), 100 units per ml penicillin/streptomycin and 2 mM glutamine, and incubated at 37°C and 5% CO 2 . Primary rat hippocampal neurons were prepared and kept in culture as described previously 40 .
Zn 2 þ loading of NG108-15 cells. Twenty-four hours after seeding, NG108-15 cells were loaded for 30 min with calcein red-orange AM (2.5 mg per well; Invitrogen Molecular Probes). Next, the cells were incubated for 4 h with one of the following solutions: (i) basal solution containing (in mM): NaCl, 140; KCl, 3; CaCl 2 , 2; MgCl 2 , 1; D-glucose, 25; HEPES/NaOH, 10 (pH 7.4); (ii) K þ solution, same as the basal solution but KCl concentration raised to 50 mM; (iii) Zn 2 þ solution, same as the basal solution but with added Zn 2 þ (200 mM); and (iv) K þ þ Zn 2 þ solution, same as the K þ solution but containing also 200 mM Zn 2 þ . Fifteen minutes after returning the cells to the DMEM incubation medium they were exposed to the fluorescent Zn 2 þ indicator N-(6-methoxy-8-quinolyl)-ptoluenesulfonamide (TSQ; 0.001% final, added from 0.5% weight per volume stock in dimethylsulfoxide). Ten minutes later, cells were examined and photographed under a fluorescence microscope (Axio Observer.A1, Zeiss). Photographs ( Â 20) were taken under identical conditions. Background-corrected calcein red-orange and TSQ fluorescence was quantified for single cells (regions of interest were set using a differential interference contrast image) using ImageJ software (NIH) and averaged for every field of view. ARTICLE synthesized by reverse transcription from total RNA using the RevertAid Premium First strand cDNA Synthesis Kit (Fermentas) following the manufacturer's protocol. Ca 2 þ channel subunit transcript quantification was performed by realtime reverse transcription-PCR (RT-PCR). Relative quantification of the starting mRNA copy numbers was carried out according to the DDC t method. The signal threshold was set within the exponential phase of the reaction for determination of the threshold cycle (C t ). PCR samples contained 1 Â Maxima Probe/ROX qPCR Master Mix (Fermentas), 5 pM each oligonucleotide primer (Supplementary  Table 2) and 1/10 synthesized cDNA in a 6.25-ml volume. Quantitative PCR was performed in an ABI Prism 7900HT apparatus (PE Applied Biosystems, Foster City, CA, USA) with conditions as follows: 2 min at 50°C, 10 min at 95°C, then 40 cycles of 15 s at 95°C and 1 min at 59°C.
Western blot analysis. For western blot analysis, cells were washed with ice-cold PBS, detached from the plates and centrifuged for 1 min at 4°C. Pellets were resuspended in PBS þ 10 mM EDTA and homogenized by sonification. Proteins were quantified using the nanodrop (ThermoScientific) and 150 mg protein was loaded on 7% Tris-glycine SDS-PAGE gels and transferred to nitrocellulose membranes. Membranes were blocked for 1 h at room temperature in 2% fish gelatin (Sigma) and then incubated for 1 h with antibodies directed against Ca V 3.2 (1:200; cat #: ACC-025; Alomone Labs) and a-tubulin (1:10,000; ab7291; Abcam). After washing and an hour incubation with the secondary antibodies (IRDye680 goat-anti-rabbit and IRDye800 goat-anti-mouse; both 1:20,000; LI-COR) in PBS þ 0.1% Tween 20, immunoreactive bands were detected using the Odyssey infrared imaging system (Li-COR Biosciences GmbH, Bad Homburg, Germany) and quantified using the AIDA software (Raytest). The a-tubulin signal was used as internal control. Full blots are shown in Supplementary Fig. 5.
Electrophysiology. Patch-clamp recordings were obtained from NG108-15 cells. The voltage dependence of activation and inactivation was characterized using standard voltage step protocols (Fig. 2a). The voltage-dependent activation of the Ca 2 þ conductance was fit by the product of a Boltzmann function, reflecting voltage-dependent activation equation (1), and the general constant field equation (2): where I Ca (V) denotes the Ca 2 þ current and g Ca (V) the Ca 2 þ conductance amplitude, respectively, at the membrane potential V as set by the command voltage V.
[Ca 2 þ ] in and [Ca 2 þ ] out correspond to internal and external Ca 2 þ concentration, respectively. The values V 1/2 (membrane potential at half-maximal inactivation or activation), A 0 and A 1 (maximal and minimal conductance, respectively) were determined by the fitting procedure. F is Faraday's constant, R is the gas constant and T is the temperature at which the measurements were conducted (22°C on average). The conductance g Ca for each potential was derived, normalized to A 0 and averaged for all cells of the same group. The voltage dependence of inactivation was derived by converting peak current to g Ca and fitting these values with equation (1).
Luciferase assay. Transfection of the NG108-15 cells was carried out using lipofectamine (Invitrogen) following the manufacturer's protocol. For each well (48-well tissue culture plates; 80% confluency), 0.5 mg Ca V 3.2 luciferase reporter plasmid, 0.0125 mg pRL-TK (Promega) and 0.5 ml lipofectamine were mixed with 25 ml serum-free medium. The mixture was incubated for 20 min at room temperature and then added to the appropriate wells. Cells were grown in serum-free culture medium at 37°C and 5% CO 2 . After 16 h, the serum-free medium was replaced by serum-containing medium and the cells were used for experiments 36 h after transfection.
Renilla luciferase was used to normalize the transfection efficiency data, and a Dual Luciferase Reporter Assay System was used according to the manufacturer's specifications (Promega). Renilla and firefly luciferase activities were determined using the Glomax Luminometer (Promega). The results are given as firefly/Renilla relative light units if not indicated otherwise.
ChIP on NG108-15 cells. NG108-15 cells (six wells; 100% confluency) were cross-linked with 1% formaldehyde for 10 min at 37°C. Cells were washed twice in cold PBS containing protease inhibitors (Complete Protease Inhibitor Cocktail Tablets; Roche), lysed in 200 ml SDS lysis buffer (1% SDS; 10 mM EDTA; 50 mM Tris, pH 8.1 with protease inhibitors) and incubated on ice for 10 min. Lysates were sonicated using an Ultrasonic Processor UP50H (Hielscher Ultrasound Technology, Germany), with four sets of 10-s pulses at 50% of maximum power. Samples were centrifuged at 13,000 r.p.m. for 10 min at 4°C and the supernatant was diluted 10-fold in ChIP dilution buffer (0.01% SDS; 1.1% Triton X-100; 1.2 mM EDTA; 16.7 mM Tris, pH 8.1; and 167 mM NaCl with protease inhibitors). Next, samples were incubated overnight at 4°C with 5 mg anti-MTF1 (C-19X, SC26844X, Santa Cruz, CA). Rabbit-IgG incubations (Cell Signaling Technology; #2729) were included as control for the immunoprecipitation. Further processing of the ChIP samples was performed using the SimpleChIP Plus Enzymatic chromatin IP kit (Cell Signaling Technology; #9005) as described above.
Viral vector production. Recombinant AAV1/2 genomes were generated by largescale triple transfection of HEK293-AAV cells. The rAAV plasmid, helper plasmids encoding rep and cap genes (pRV1 and pH21), and adenoviral helper pFD6 (Stratagene, La Jolla, USA) were transfected using standard CaPO 4 transfection. Cells were collected B60 h following transfection. Cell pellets were lysed in the presence of 0.5% sodium deoxycholate (Sigma) and 50 U ml À 1 Benzonase endonuclease (Sigma). rAAV viral particles were purified from the cell lysate by HiTrap heparin column purification (GE Healthcare), and then concentrated to a final stock volume of 400 ml using Amicon Ultra Centrifugal Filters (Millipore). Purity of the viruses was validated by coomassie blue staining of SDS-polyacrylamide gels. Functional titres (transducing units) of the fluorescent protein vectors were determined by transduction of cultured primary neurons.
Animal experiments. Infusion of AAV vectors. Mice and rats were housed under a 12 h light/dark cycle with food and water ad libitum. All experiments were performed in accordance with the guidelines of the European Union and the University of Bonn Medical Center Animal Care Committee. Adult male mice (B50 days, 420 g) were obtained from Charles River (C57Bl/6-N) and were anesthetized with 6 mg kg À 1 xylazine (Rompun; Bayer) plus 90-120 mg kg À 1 ketamine, intraperitoneal (i.p.) (Ketavet; Pfizer). Intracerebral injection of viral particles in the left and right CA1 hippocampal region was performed stereotactically at the coordinates (in mm) À 2 posterior, À 2/2 lateral and 1.7 ventral relative to bregma. Holes the size of the injection needle were drilled into the skull, and 1 ml of viral suspension containing B10 8 transducing units was injected using a 10 ml Hamilton syringe at a rate of 100 nl min À 1 using a microprocessor-controlled mini-pump (World Precision Instruments). After injection, the needle was left in place for 5 min before withdrawal. The needle was then slowly withdrawn and the incision closed.
Near-infrared in vivo imaging. MTF1 overexpression. Mice were injected with rAAV-Ca V 3.2-iRFP particles as described above. Two weeks after injection, mice were anesthetized, the skull was exposed and holes were drilled at the same location for subsequent rAAV-Syn-MTF1-IRES-Venus or rAAV-Syn-Venus injection. Just before injection, basal iRFP values were measured through the skull. Three weeks after injection of viruses harbouring MTF1-or Venus-expressing constructs, iRFP values were again determined. ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms9688 Zn 2 þ injection. Animals were injected with rAAV-Ca V 3.2-iRFP particles in the hippocampal CA1 region and analysed for their basal iRFP levels 3 weeks after injection. Next, they were injected with 1 ml 100 mM ZnCl 2 or 1 ml 0.9% NaCl in the same holes as used for the rAAV particles. Animals were imaged again 3 days after injection.
Near-infrared imaging was performed with a Pearl Impulse Small Animal Imaging System (Li-COR Biosciences GmbH). The iRFP signal was determined using a highly sensitive charged-coupled device camera. Excitation and emission wavelengths were fixed at 690 and 710 nm, respectively. Pictures were analysed using the Pearl Impulse Image Studio Software v3.1 (Li-COR Biosciences GmbH). Fluorescent signals were normalized to background levels and quantified by placing two round regions of interest (Fig. 6e) above the hippocampal region 39 . Fluorescent signals are presented as arbitrary units (a.u.).
At the end of the in vivo imaging experiments, mice were decapitated under deep isoflurane (Forene) anesthesia. Brains were removed and fixed in formaldehyde overnight. Coronal brain slices (30 mm) were made on a vibratome (Leica), mounted on slides (Histobond, Marienfelde Germany) and imaged on the Odyssey infrared imaging system (Li-COR Biosciences GmbH). Only animals with hippocampal CA1 staining (Fig. 6f) were included in the in vivo imaging analysis.
EEG-video monitoring. The electrographic features of Venus-and MTF1DCinjected animals after pilocarpine-SE were analysed with a telemetric EEG/videomonitoring system (Data Science International) as described previously 13,41 . Mice were implanted with EEG electrodes directly after the viral injection (described above), when the animals were still narcotized. The transmitter was placed subcutaneously on the right abdominal side and stainless screws were used as electrodes and positioned in the two holes used for the viral injection (2 mm posterior and 2 mm lateral to Bregma). EEG recording with a sampling rate of 1 kHz was started directly after the implantation procedure. Two weeks after the implantation, animals experienced pilocarpine-SE. Monitoring was performed continuously until day 20 after SE. EEG recordings were analysed using NeuroScore v2.1 (DSI) software with the following parameters: a threshold value between 100 and 5,000 mV; a spike duration between 5 and 70 ms; a spike interval between 0.005 and 1 s; a minimum train duration of 5 s; and a minimum number of spikes per event of 10. These parameters were verified by analysing two animals manually for the whole timeframe. All positive events (seizures) of the two animals were identified using these parameters. The output of all animals was checked manually for false-positive events (for example, artefacts) and all false-positive events were deleted from the output files. From concurrent video recordings, all seizures were classified as described previously 41 .
Adult male Sabra rats (150-200 g) were kept in the animal facility of Hebrew University and Hadassah School of Medicine. All the experiments were approved by the local institution's ethical committee. Rats were injected with scopolamine methyl nitrate (1 mg kg À 1 , s.c; Sigma). Thirty minutes afterwards, SE was induced with a single dose of pilocarpine (350 mg kg À 1 (i.p.); Sigma). SE was terminated after 2 h by diazepam (Ratiopharm; 0.1 mg kg À 1 (i.p.)).
Zn 2 þ imaging in rat brain slices. Pilocarpine-injected rats were decapitated under isoflurane anesthesia at different time points after the termination of SE. The brains were rapidly extracted and immediately frozen in liquid nitrogen and were stored at À 80°C. Either transversal (2-5.5 mm below the inter-auricular plane, containing the ventral part of the hippocampus) or saggital (1-3.5 mm laterally to the inter-hemispheric plane, containing the dorsal hippocampus) slices, 20-mm wide, were cut in a cryotome and placed on glass slides. For the staining we used 1 of every 10 slices, 200 mm apart from each other, for a total of 15 transverse and 12 sagittal slices for each hemisphere.
For visualization of chelatable Zn 2 þ , fresh frozen brain slices were examined under a fluorescent microscope (Olympus BX60, Germany; excitation filter 330-385 nm, emission filter 420 nm), at least 30 min after the preparation. TFL-Zn (Sigma), 250 mM in saline, was applied to the slices, immediately before the microscopic examination, for 30 s and washed with saline. Stained CA1 hippocampal cells were counted by use of a gridded microscope lens (with calculated area of 0.05 mm 2 in high-power field). The density of Zn 2 þ -stained neurons was defined as the number of Zn 2 þ -stained neurons counted in a 0.05-mm 2 area of the slice (grid surface) at high-power field.
Human TLE patients and mRNA expression analyses. For gene expression analyses, we used human hippocampal biopsy tissue from patients with hippocampal sclerosis (n ¼ 79) versus patients with lesion-associated (low-grade neoplasms or dysplasia; n ¼ 35) chronic TLE, who underwent surgical treatment in the Epilepsy Surgery Program at the University of Bonn Medical Center due to pharmacoresistance. In all patients, presurgical evaluation using a combination of non-invasive and invasive procedures revealed that seizures originated in the mesial temporal lobe 42 . All procedures were conducted in accordance with the Declaration of Helsinki and approved by the Ethics Committee of the University of Bonn Medical Center. Informed written consent was obtained from all patients. Clinical characteristics per subgroup are described in Supplementary Table 3. mRNA analyses for Ca V 3.2 and MTF1 were carried out analogous to a procedure described elsewhere in detail 43 . Briefly, RNA from biopsies representing all hippocampal subfields served to generate 750 ng cRNA used for hybridization on Human HT-12 v3 Expression BeadChips with Illumina Direct Hybridization Assay Kit (Illumina, San Diego, CA) according to standard procedures. We extracted data for Ca V 3.2 and MTF1 analysed by Illumina's GenomeStudio Gene Expression Module and normalized using Illumina BeadStudio software suite by quantile normalization with background subtraction.
Statistical analysis. Statistical analyses were performed with GraphPad Prism 6.05 software (GraphPad Software). Sample size (n) per experiment was calculated using power analysis, with parameters set within the accuracy of the respective experiment. Student's t-tests and repeated measures analysis of variance followed by Bonferroni's multiple comparisons or Tukey's multiple comparisons tests were used to evaluate the statistical significance of the results. Values were considered significantly at Po0.05. All results are plotted as mean±s.e.m. All electrophysiological and animal experiments were conducted in a randomized and blinded fashion. All in vitro experiments were independently repeated at least two times.