Aberrant expression of PAR bZIP transcription factors is associated with epileptogenesis, focus on hepatic leukemia factor

Epilepsy is a widespread neurological disease characterized by abnormal neuronal activity resulting in recurrent seizures. There is mounting evidence that a circadian system disruption, involving clock genes and their downstream transcriptional regulators, is associated with epilepsy. In this study, we characterized the hippocampal expression of clock genes and PAR bZIP transcription factors (TFs) in a mouse model of temporal lobe epilepsy induced by intrahippocampal injection of kainic acid (KA). The expression of PAR bZIP TFs was significantly altered following KA injection as well as in other rodent models of acquired epilepsy. Although the PAR bZIP TFs are regulated by proinflammatory cytokines in peripheral tissues, we discovered that the regulation of their expression is inflammation-independent in hippocampal tissue and rather mediated by clock genes and hyperexcitability. Furthermore, we report that hepatic leukemia factor (Hlf), a member of PAR bZIP TFs family, is invariably downregulated in animal models of acquired epilepsy, regulates neuronal activity in vitro and its overexpression in dentate gyrus neurons in vivo leads to altered expression of genes associated with seizures and epilepsy. Overall, our study provides further evidence of PAR bZIP TFs involvement in epileptogenesis and points to Hlf as the key player.

Normalization was performed using the RMA normalization using all probes regardless of Affymetrix 'Present' (P) or ' Absent' (A) calls. Only probes with the "at" suffix were retained to ensure that only probes specific to a single gene were carried forward for further analysis. Differential expression of the genes Hlf, Tef, Dbp and E4bp4 was performed using the Limma (Linear models for microarray) package 27  www.nature.com/scientificreports www.nature.com/scientificreports/   www.nature.com/scientificreports www.nature.com/scientificreports/ Devices). Stimulation of neurons was achieved by adding bicuculline (20 μM) to the recording bath, while delivering glutamate (2.5 mM) with a patch pipette with a resistance of ~2MΩ placed near the recorded cell (~50μm away). Glutamate concentration in the pipette was 2.5 mM, and delivery of glutamate was obtained by a 50 μl Hamilton syringe mounted to a SP101i syringe pump (flow rate of 1 μl/min, World Precision Instruments, Fl, USA) and connected to the pipette by a tight silicon tubing. Action potentials were obtained by applying a series of current steps (50 pA each) under current clamp mode. Traces were recorded with pClamp 10 and analysis of intrinsic properties and sPSCs were performed with MiniAnalysis (Synaptosoft). Average traces were obtained by selecting events in MiniAnalysis, aligning the peak of the events and averaging those traces under the "group analysis" function. Average traces from each cell were averaged to obtain one trace per group. Decay fit was calculated in Mini Analysis by using a 2 exponential function approach (Amplitude1 exp(−time/τ1) + Amplitude2 exp(−time/τ2).

TNF-
Two-way ANOVAs and two-tailed t-tests were performed with GraphPad Prism version 8 (GraphPad Software). The amplitude and interevent interval of PSCs were analyzed according to the highest level of activity shown, in one-minute long traces. To make the KS tests more stringent we applied a bootstrap estimation of the p-value by extracting random samples of size 100 (either currents or amplitudes) using the function "sample" on R statistical package 25 . After obtaining a p-value from these samples the procedure was repeated 1000 times. The percentage of times the random samples gave a p-value < 5% was used to establish if cumulative distributions are different. We considered distributions to be significantly different when more than 50% of random samples had a p-value < 5%This ensures that the KS test is not as sensitive to small differences between the associated functions. Density plot of sPSCs amplitudes and interevent intervals were used to visually confirm our statistical results.

transcriptomics analysis. Next-generation sequencing was performed by Functional Genomics Center
Zurich, University of Zurich, Switzerland. Consequent data analysis was than performed by the Swiss Institute of Bioinformatics, Switzerland. Total RNA was processed using the TruSeq RNA stranded protocol (Illumina) in order to produce sequencing libraries. Libraries were then sequenced on the Illumina HiSeq. 4000, single-end, 125 cycles. Reads were aligned against Mus Musculus.GRCm38.86 genome using STAR (v.2.5.2b) 29 . The number of read counts per gene locus was summarized with htseq-count (v0.6.1) 30 using Mus Musculus.GRCm38.86 gene annotations. Quality of the RNA-seq data alignment was assessed using RSeQC (v. 2.3.7) 31 . Reads were also aligned to the Mus Musculus.GRCm38.86 transcriptome using STAR (v. 2.5.2b) 29 and the estimation of the isoforms abundance was computed using RSEM (v. 1.2.31) 32 . To assess differential expression, we used the R Bioconductor package DESeq. 2 (version 1.14.1) 33 . Differentially expressed Genes were identified at the Benjamini-Hochberg (BH) adjusted P < 0.05 level, using the Wald test. For gene set enrichment analysis, no direction criterion on fold change was applied. Enriched GO (Gene Onthology) categories were identified using the enrichment analysis package in R/Bioconductor, clusterProfiler 34 with significant terms at BH P adjusted <0.05 reported.

Status epilepticus induces changes in the expression of pAR bZip transcription factors and
does not affect expression of clock core genes. We evaluated the hippocampal expression of core clock genes and PAR bZIP TFs at different stages during epileptogenesis by injecting KA into the right dorsal hippocampus, a well-established model of TLE 35 . The pathogenesis of this model follows a stereotypic pattern 23,35 , characterized by an initial status epilepticus lasting up to 1 dpi, followed by a silent latent period and the occurrence of spontaneous recurrent seizures in the chronic period starting at about 14 dpi. The expression of clock genes and PAR bZIP TFs was evaluated at two (6 and 14 dpi) or four ZT time-points (1 and 28 dpi) over a day to exclude possible phase shift in clock gene expression amplitude. We found significant alteration in the expression of PAR bZIP TFs (Fig. 2). The transcriptional activators Hlf, Dbp and Tef were significantly down-regulated while the transcriptional repressor E4bp4 was up-regulated. Changes were most evident during the acute and epileptogenesis phases (1 and 6 dpi, respectively) of the status-epilepticus and tend to normalize thereafter. The expression of Hlf remained significantly down-regulated also at the beginning of the chronic phase, 14 days after the induction of status epilepticus. At 28 dpi, Hlf was significantly suppressed only at ZT18. Interestingly, similar changes in expression of PAR bZIP TFs were found in the contralateral (not-injected) hippocampus at 24 h post injection. Additionally, Hlf remained significantly downregulated during the later chronic period (28 dpi) in the contralateral hippocampus (Fig. 3). Surprisingly, we did not observe any significant changes in the expression of the following clock genes: Clock, Bmal1, Npas2, Per1, Per2, Cry1, Cry2, Rev-erb-α and Rorc (Fig. 4). Only the expression of Per3 was downregulated 24 h after the induction of status epilepticus (Fig. 4). In the contralateral hippocampus, there were no significant changes in the expression of core clock genes except downregulation of Per3 at 1 dpi, ZT0 and 6 (see Supplementary Fig. 1). To ensure that the observed changes are due to the local KA injection, we analyzed in addition the expression of PAR bZIP TFs and core clock genes in the cortical tissue above the injected hippocampus. We did not detect any significant changes in this tissue (see Supplementary Fig. 2).
To evaluate whether changes in PAR bZIP TFs can be observed in other rodent models of TLE, we performed analysis on previously published 26 microarray data (NCBI Gene Expression Omnibus. GSE47752) profiling the hippocampal expression of genes in four different rat models of acquired epilepsy: kainate (intraperitoneal, i.p.), pilocarpine, self-sustained status epilepticus (SSSE) and amygdala kindling. We focused our analysis on the differential expression of PAR bZIP factors in these models at 24 h after induction of status epilepticus. Our analysis revealed signification downregulation of Hlf in all four models (See Fig. 5a). In addition, Tef was significantly downregulated and the transcriptional repressor E4bp4 was significantly up-regulated in all models except for SSSE. Dbp was significantly downregulated in the pilocarpine model. Taken together the microarray data of four rat models of epilepsy support our finding of altered expression of PAR bZIP TFs in KA induced epilepsy. (2020) 10:3760 | https://doi.org/10.1038/s41598-020-60638-7 www.nature.com/scientificreports www.nature.com/scientificreports/ excitotoxicity and seizure activity induce the downregulation of Hlf. Next, we investigated whether acute hyperexcitability in general is responsible for the altered behavior of PAR bZIP TFs. To explore this possibility, we injected either N-methyl-D-aspartic acid (NMDA) or bicuculline into the right dorsal hippocampus. Hippocampal injection of NMDA stimulates ionotropic NMDA receptors, induces acute excitotoxicity and status epilepticus; however, unlike KA-induced status epilepticus, does not cause an early loss of hippocampal interneurons 22,36 . Bicuculline is a competitive antagonist of GABA A receptors and it is known to produce seizures in the absence of neurodegenerative events 37 . In both models, Hlf was significantly downregulated 24 h after status epilepticus (Fig. 5b), whereas Dbp was upregulated only after NMDA injection. These effects were mirrored also in the contralateral hippocampus (Fig. 5b).

Changes in the expression of PAR bZIP transcription factors are independent of inflammation.
Both clinical [38][39][40] and experimental 20,41,42 evidence suggest that inflammatory processes are involved in the pathogenesis of TLE. Hence, we characterized the expression of proinflammatory cytokines TNF-α and IL-1β, as well as the anti-inflammatory cytokine interleukin 10 (IL-10) in the intrahippocampal KA model of TLE (Fig. 6). TNF-α expression was significantly upregulated 1, 6 and 14 days after the KA infusion (15-fold, 10-fold and 5-fold respectively). IL-1β beta remained upregulated at all timepoints studied, the effect being about 5-fold. The expression of IL-10 was increased at later time points, at 14 and 28 dpi. These cytokines were also significantly upregulated in the contralateral hemisphere, albeit the increase being less pronounced (Fig. 6).
As outlined above, inflammation regulates the expression of PAR bZIP TFs. Thus, we hypothesized that the KA induced inflammatory response drives the observed changes of PAR bZIP TFs expression. To test this hypothesis, we infused either mouse cytokine TNF-α or IL-1β into the hippocampus. Whereas TNF-α mRNA expression did not increase upon TNF-α or IL-1β injection, the expression of IL-1β mRNA was augmented about 5-and 10-fold respectively (Fig. 7a). However, we observed no changes in the expression of PAR bZIP factors (Fig. 7a). As cytokine levels reached by this treatment may have been insufficient to inhibit PAR bZIP TFs, we decided to www.nature.com/scientificreports www.nature.com/scientificreports/ simulate sub-chronic TNF-α exposure by infusion of an adenovirus-associated viral TNF vector (AAV-TNF). The AAV-TNF induced robust TNF-α expression (100-fold) in the hippocampus; however, it did not affect the expression of PAR bZIP TFs (Fig. 7b). We corroborated this result by inducing neuroinflammation by infusion of lipopolysaccharide (LPS). LPS activates microglia interacting with their Toll-like receptor-4 and induces production of inflammatory cytokines. The intrahippocampal infusion of LPS induced significant overexpression of TNF-α (50-fold) and IL-1β (100-fold); however, the expression of PAR bZIP TFs remained unchanged (Fig. 7c). Surprisingly, the hippocampal expression of PAR bZIP TFs was not affected by a local increase in the expression of pro-inflammatory cytokines such as TNF-α and IL-1β, as has been described for other murine tissues 15 . This evidence is in line with our previous finding that the expression of PAR bZIP TFs is unchanged in brains of mice with experimental autoimmune encephalomyelitis, a model of multiple sclerosis (unpublished data). Hence, this evidence suggests that early changes in hippocampal expression of PAR bZIP TFs in rodent models of TLE are not a direct response to inflammatory processes that occur during the initial stages of epileptogenesis. Thus, PAR bZIP TFs in the hippocampal tissues are regulated in a different way than in the periphery.

KA induces early response of clock genes.
To further explore the mechanism responsible for the downregulation of PAR bZIP TFs during epileptogenesis, we decided to characterize the hippocampal expression of PAR bZIP TFs and core clock genes, at 3 hours after the induction of status epilepticus by intrahippocampal KA (Fig. 8). At this early time point, we observed significant reduction of Clock, Bmal1, Npas2 and Per3 expression. Additionally, the expression of Per1 and Per2 was significantly upregulated. The expression of PAR bZIP TFs Hlf and Dbp did not show statistically significant decrease, while the expression of Tef was significantly downregulated. This expression pattern suggests that early changes in core clock genes might be involved in the initiation of changes in PAR bZIP TFs levels. The positive regulators of their expression (Clock, Bmal1, Npas2) were downregulated and their negative regulators (Per1 and Per2) were upregulated. Additionally, the expression of TNF-α www.nature.com/scientificreports www.nature.com/scientificreports/ and IL-1β was not altered yet at this early time point. These results might explain the changes in PAR bZIP TFs expression at 1 dpi after the KA lesion.
neurons over-expressing Hlf showed a significant decrease in the frequency and increase in the amplitude of spontaneous events in the presence of bicuculline and glutamate. As we observed consistent downregulation of Hlf expression in various models of epilepsy, likely as a result of acute hyperexcitability, we examined the effects of Hlf expression in neurons under basal and hyperexcitable conditions. Hence, we overexpressed Hlf by transducing primary hippocampal neurons at 2 DIV with either a control vector (AAV-EGFP) or a vector overexpressing Hlf (AAV-HLF) and recorded spontaneous currents at ~12-16 DIV by whole-cell voltage clamp. Under basal activity levels (Fig. 9a,b,c), overexpression of Hlf resulted in a significant reduction in the amplitude of sPSCs, with no differences in the frequency or decay kinetics (Fig. 9d, AAV-EGGP: n = 10, 8.72 ± 0.89 msec; AAV-HLF: n = 11, 8.01 ± 1.06 msec, p = 0.6204), of events compared to neurons transduced with an AAV-EGFP. In the presence of bicuculline and pressure-ejected glutamate (Fig. 9e,f,g), neurons overexpressing Hlf showed a significant decrease in the frequency and an increase in the amplitude of sEPSCs compared to neurons carrying the EV control. We observed no differences between groups in the number of action potentials [AAV-EGFP: n = 10, 5.     www.nature.com/scientificreports www.nature.com/scientificreports/ Gene expression analysis. To determine which genes are regulated by hepatic leukemia factor, we performed transcriptome analysis of the dorsal hippocampal tissue from mice over-expressing mouse Hlf restricted to the dentate gyrus (Fig. 1a). It has been shown that that mouse endogenous Hlf is expressed mostly in the dentate gyrus, while in CA1 and CA3 subregions of hippocampus the expression is limited 43 , (Experiment: 565 and 566, Probe Name: Rp_Baylor_103366) 44 . These samples were compared with samples from mice injected with a control vector (AAV-EGFP). Hlf was over-expressed 7-fold in the dorsal hippocampal tissue while there were no changes in the expression of Tef and Dbp (Fig. 1b). In addition, the AAV-HLF injection did not induce www.nature.com/scientificreports www.nature.com/scientificreports/ local neuroinflammation detected as expression of TNF-α and IL-1β (Fig. 1b). The expression of AAV-HLF vector was restricted to neurons and was detected in neither astrocytes nor microglia (Fig. 1c). Only samples with confirmed overexpression of Hlf using qPCR were sequenced (AAV-HLF n = 7, AAV-EGFP n = 7). In total, the expression of 65 genes was significantly altered (Fig. 1d). The list of all differentially expressed genes is included in the Supplementary Table S1. Raw data are available at NCBI GEO database under accession number GSE140046. Additionally, the gene ontology classification for molecular function and biological process were performed using all 65 significantly differentially expressed genes (Fig. 1e,f).
We identified genes whose expression is influenced by neuronal Hlf. We found changes in expression of genes coding for ion channels such as Trpa1, P2rx5, Grin3a, Kcng1 and Gabrd. Next, the expression of Synpr, Gfra2, Lcn2, Slc30a3, Rasd2, Igfbp5, Fxyd7, Cdkn1a and Slc6a8 was altered. Synaptoporin, the protein coded by Synpr gene, is a marker of mossy fiber sprouting, a phenomenon associated with epileptogenesis after KA injection 45 . Gfra2 was reported to modulate threshold of kindling evoked seizures 46 , Lcn2 was identified as a chemokine inducer in KA model 47 and Slc30a3 modulates transport of zinc into synaptic vesicles that is co-released with glutamate and regulates excitability and its loss is associated with febrile seizures 48 . Rasd2 and Igfbp5 were identified by a recent study as candidate genes associated with epileptogenicity in KA model of TLE 49 and Fxyd7 to be altered after seizure preconditioning using pilocarpine model 50 . Human studies reported Cdkn1a to be upregulated in patients with epilepsy 51 and deficiency in Slc6a8 was reported to result in intractable epilepsy and cognitive impairment 52 . Although, with the given dataset, we are not able to deduce any direct involvement in seizure modulation, many of those genes were previously described to be associated with neuronal excitability, seizures or epilepsy.

Discussion
Even though the occurrence of epileptic seizures in circadian patterns has been documented in animal models as well as humans 53 the relationship between epileptic seizures and the circadian system at the molecular level is still unclear. A recent study reported decreased Clock expression in human epileptogenic tissue and decreased seizure www.nature.com/scientificreports www.nature.com/scientificreports/ threshold in mice with Clock deletion in excitatory cortical neurons suggesting that alterations in Clock expression are epileptogenic 3 . Furthermore, mice deficient for Bmal1 exhibited reduced seizure thresholds 13 .
In this study, we utilized the KA model of TLE to study the expression of core clock genes and their downstream transcription factors, the PAR bZIP TFs family. We showed that the expression of Hlf, Dbp, Tef is suppressed while the expression of E4bp4 is significantly upregulated during epileptogenesis. This effect was most evident during the acute phase (1 dpi); however, it was significant also during epileptogenesis (6 dpi) and at the beginning of the chronic phase (14 dpi). Additionally, our analysis of existing microarray data set (Dingledine R. 2013. NCBI Gene Expression Omnibus. GSE47752) revealed identical trends in the expression of PAR bZIP TFs in rat models of acquired epilepsy. Surprisingly, the expression of core clock-genes was not altered at 1, 6, 14 or 28 dpi after KA injection. These data are in accordance with another study performing gene profiling in dorsal hippocampus after 6 hours, 15 days and 6 months post KA injection 54 . They did not find significant changes in core-clock gene expression while they identified Hlf to be an important transcription factor associated with changes in gene expression at 15 dpi. The functional importance of PAR bZIP TFs in regulating neuronal excitability was proven by a study showing that triple-knockout mice deficient for Hlf, Dbp and Tef develop audiogenic and spontaneous seizures 12 . The authors pointed at Tef as the gene associated with seizures by regulating expression of pyridoxal kinase, an enzyme involved in the metabolism of neurotransmitters. In another study, the deletion of Hlf in mouse exacerbated seizures and reduced survival in the Snc2a Q54 mouse model of epilepsy 55 . Likewise, DBP was found downregulated together with CLOCK, CRY1 and PER1 at protein levels in patients with focal cortical dysplasia 3 . Furthermore, neuroprotective effects were described for E4bp4, which we found here to be upregulated after induction of status epilepticus. Besides modulating circadian rhythms, E4bp4 serves as a survival factor in motoneurons during their development 56 . By limiting neuronal injury, the upregulation of E4bp4 in experimental epilepsy may be a part of the response to neurotoxic insults.
Interestingly, KA induced changes in the expression of PAR bZIP TFs as well as neuroinflammation were mirrored in the contralateral (non-injected) hippocampus. Hlf remained downregulated there even at the late chronic phase (28dpi). At the acute phase, this effect is most likely attributed to hyperactivity spreading from the injected hippocampus via commissural projections 57 . However, during the chronic phase the epileptic activity does not spread into the contralateral hippocampus 58 . Additionally, it has been shown that contralateral hippocampus can provide important regulatory input and modulate ongoing seizures 59 . Thus, it is not clear whether the downregulation of Hlf in contralateral hippocampus during the chronic phase is caused by ongoing changes in the ipsilateral hippocampus or it is entrained by intrinsic changes in the contralateral hippocampus induced during the status epilepticus.
What induces the changes in PAR bZIP TFs expression? We have shown in our previous research that the expression of PAR bZIP TFs is suppressed by the proinflammatory cytokines TNF-α and IL-1β by interfering with their E-boxes in mouse liver and fibroblasts 15 . Besides suppressing Hlf, Dbp and Tef, TNF-α was found to enhance the expression of E4bp4 in synovial fibroblasts 18 . This picture is identical to the one detected here in our KA induced model of epilepsy. Since we observed an elevated expression of TNF-α and IL-1β in the KA model, we thought that the cytokine induced neuroinflammatory effect might be responsible for the modulation of PAR bZIP TFs. However, we were not able to suppress the expression of PAR bZIP TFs in hippocampus by simulating the proinflammatory environment. This suggests that their expression in the hippocampus is regulated by mechanisms than differ from those in periphery. Distinct circadian oscillations of PAR bZIP TFs in different tissues supports this hypothesis. In liver and kidney, PAR bZIP TFs oscillate diurnally with a high amplitude 60,61 while in brain tissue they oscillate with a low amplitude or not at all 11,12 .
Early changes in clock core-genes expression might be driving the initial downregulation of PAR bZIP TFs. At an early time-point (3 h) after KA injection, when the inflammatory cytokines are not yet elevated, the expression of Clock, Npas2 and Bmal1 were significantly downregulated while Per genes 1 and 2 were upregulated. The expression of Dbp and presumably also Hlf and Tef is directly activated by CLOCK-BMAL1 62,63 and inhibited by PER and CRY 64 , thus their deficient expression results in altered expression of PAR bZIP TFs. In addition, it has been shown that Hlf and Tef expression in neurons is reduced in mouse with neuronal Clock deletion 3 . However, it is unclear whether these early changes in core clock gene expression contribute to alterations in PAR bZIP TFs at later stages of KA model, since their expression is already normalized at this stage. Alternatively, the neuroinflammation, excitotoxicity or seizure activity itself might play a role in sustaining the altered levels of PAR bZIP TFs.
To further investigate the role of PAR bZIP TFs in the regulation of neuronal excitability, we induced Hlf over-expression in primary neurons under hyperexcitable conditions induced by glutamate and bicuculline exposure. It resulted in decrease in the frequency and an increase in the amplitude of sEPSCs. In another study, Hlf deletion has been shown to increase the frequency of spontaneous seizures in two mouse models involving mutations on voltage-gated sodium channels; the gain-of-function Scn2a Q54 (Q54) mouse model and the heterozygous loss-of-function Scn1a mouse model 55 . Interestingly, these models seem to induce seizures from different mechanisms; a reduction in GABAergic inhibition in the Scn1a model and an increase in excitability in CA1 glutamatergic neurons in the Q54 mouse model 65 . This posits the possibility that Hlf is involved in regulation of both pathways, which may explain the lower excitability we observed when exposing hippocampal primary neurons to hyperexcitable conditions. This decrease in excitability does not appear to involve changes in intrinsic properties of neurons nor in dendritic filtering, as we did not observe changes in the number or amplitude of action potentials, nor in the decay time of spontaneous currents. This result suggests that Hlf, as a member of PAR bZIP TFs, is an intrinsic regulatory element modulating synaptic activity by a yet unknown mechanism.
Taken together, Hlf seems to be a promising transcription factor associated with epileptogenesis. We found consistent deficiency of Hlf in animal models of acquired epilepsy, demonstrated that Hlf regulates neuronal activity and its overexpression in neurons leads to altered expression of genes associated with epilepsy. Upon resolving the molecular mechanism and causality between PAR bZIP TFs alterations and seizures, these findings should initiate further studies into PAR bZIP TFs as a target to prevent epileptogenesis or to modulate seizure activity.