mTOR inhibitor Treats Autistic-like Behaviors in Tsc2+/- rats but not following Developmental Status Epilepticus

Background: Tuberous Sclerosis Complex (TSC), a multi-system genetic disorder often associated with Autism Spectrum Disorder (ASD), is caused by mutations of TSC1 or TSC2, which lead to constitutive overactivation of mammalian Target of Rapamycin (mTOR). In several Tsc1+/- and Tsc2+/- animal models, cognitive and social behavior decits were reversed by mTOR inhibitors. However, phase II studies have not yet shown amelioration of ASD and cognitive decits in individuals with TSC during mTOR inhibitor therapy. We asked here if developmental epilepsy, common in the majority of individuals with TSC, but absent in most animal models, could explain the discrepancy. Methods: At postnatal day P12, developmental status epilepticus (DSE) was induced in male Tsc2+/-(Eker) and wild-type rats, establishing four experimental groups including controls. In adult animals (n = 36), behavior was assessed in the paradigms of social interaction test, elevated plus-maze, light-dark test, Y-maze and novel object recognition. Electroencephalographic (EEG) activity was recorded in a separate set of animals (n = 18). The testing was carried out before medication (T1), during treatment with the mTOR inhibitor everolimus (T2) and after washing-out (T3). Results: Both Tsc2+/- mutation and DSE caused social behavior decits and epileptiform EEG abnormalities (T1). Everolimus led to persistent improvement of the social decit induced by Tsc2+/-, while decits related to DSE did not respond to everolimus (T2, T3). Conclusions: These ndings may contribute to an explanation why ASD symptoms in individuals with TSC, where comorbid early-onset epilepsy is common, were so far in clinical studies not reliably ameliorated by mTOR inhibitors.


Background
Autism spectrum disorder (ASD) is a neurodevelopmental disorder occurring in about 1% of the general population and well recognized as a global public health concern [1]. At least 25% of people with ASD have epilepsy, associated with developmental delay, a range of problem behaviors and signi cant treatment challenges [2]. However, there is still a very limited understanding of the fundamental links between ASD and seizures and a dearth of pre-clinical and clinical intervention studies of ASD in the context of epilepsy.
Tuberous Sclerosis Complex (TSC) is an autosomal dominant disorder with an incidence of 1:6000, with manifestations (including, but not limited to, benign tumors and other types of lesions) that can affect almost every organ in the body, including the brain. TSC is associated with ASD in up to 50% of individuals, and in turn, accounts for 1-4% of overall autism cases [3][4][5]. A lifetime history of epilepsy is reported in 70-90% of individuals with the disorder [6].
TSC is caused by mutations in either TSC1 or TSC2 resulting in dysfunction of the TSC1-TSC2 intracellular protein complex, causing overactivation of the mTOR signaling pathway [7]. The pharmacological class of mTOR inhibitors (mTORi) have emerged as molecularly targeted treatments for TSC1-TSC2-protein complex overactivation in TSC. For several organ manifestations in TSC, including renal angiomyolipoma, subependymal giant cell astrocytoma (SEGA) and as adjunctive treatment in treatment-resistent epilepsy, mTORi are now FDA and EMA approved therapies [8][9][10][11].
Apart from ASD, TSC is associated with a wide range of TSC-associated neuropsychiatric disorders (TAND) seen in up to 90% of individuals with TSC [7]. There is therefore signi cant interest in understanding the molecular (and other) mechanisms of TAND in order to identify appropriate treatments. After earlier proposals that ASD in TSC was caused by structural and/or electrophysiological aberrations, de Vries and Howe [12] proposed that direct molecular pathways may be su cient to lead to ASD and other TAND de cits, and that mTORi may therefore be molecular targeted treatments [12].
In Tsc1 and Tsc2 animal models, mTORi were shown to reverse cognitive [13] and social behavior impairments [14,15], and early-phase mTORi in adults with TSC suggested potential improvement in memory and executive skills [16]. However, in subsequent trials of children and adolescents, two phase II trials reported no signi cant improvement in TAND manifestations including intellectual ability, behavioral problems or ASD-related symptoms, after 6-12 months of mTORi administration [17,18].
We previously reported social de cits in adult Eker rats (a spontaneous Tsc2 haploinsu ciency model) reminiscent of the ndings in Tsc1+/and Tsc2+/mice. We also observed social de cits induced by early developmental status epilepticus (DSE) in wild-type rats, and reported additional social de cits in Tsc2+/rats when combined with DSE [30][31][32]. The combination of Tsc2+/and DSE corresponds more closely to the situation of the majority of individuals with TSC plus ASD. Recently, we reported that social de cits in the combined Tsc2+/plus DSE model responded to the mTORi everolimus [33]. However, whether the autistic-like behaviors induced by Tsc2+/-, by DSE or by combination of TSC and DSE have differential responsiveness to mTORi has not been examined to date.
Taking together the clinical observations and our experimental animal work, we hypothesized that ASD phenotypes in TSC resulted from a combination of TSC1/TSC2 molecular de cits (directly leading to some social de cits) and additional social impairments caused by seizures. We proposed that the direct TSC molecular de cits may be sensitive to mTORi, but that seizure-related de cits may not be, thus explaining why TAND manifestations in individuals with TSC may have been unresponsive to mTORi. Here, we studied the effect of everolimus, an mTORi, on the social de cits induced by the two factors (Tsc2+/-and DSE), either in isolation or in combination.

Animals
Rats used in this study were heterozygous Tsc2+/-(Eker) males, RRID:RGD_625624 (homozygous mutants are not viable) and their wild-type Long-Evans littermates, bred at the National Institute of Mental Health (NIMH) in Klecany, Czech Republic. Breeding Tsc2+/males originated from the breeding colony at Technische Universität Dresden, Germany, or from subsequent generations born at NIMH. Long-Evans dams, RRID:RGD_2308852, were purchased from a supplier (Charles River, Germany). We took care to use wild-type females only, as dam genotype was demonstrated to in uence maternal care and pup behavior in Tsc2+/mice [34]. Breeding was performed in individually ventilated cages (Tecniplast, 40 × 35 × 21 cm) and after weaning at postnatal day (P)28, the offspring was transferred to standard Plexiglas boxes (44 × 28 × 23 cm) in an air-conditioned room with a 12 h/12 h light-dark cycle and food and water ad libitum.
On P3-P4, pups were sexed and genotyped by PCR using tissue samples from tail tips [35]. The offspring showed the expected Mendelian ratio 1:1 for Tsc2+/and wild-type genotypes.
For behavioral experiments, we used males (n = 36) from ve different litters (littermates evenly distributed among experimental groups for counterbalanced design). Additional adult Long-Evans males from different litters, which underwent the same treatment as the main group, were used for the EEG testing (n = 18) and everolimus quanti cation (n = 16). Individual animals were arbitrarily distributed to experimental groups. Each rat had a unique identi cation code independent on group membership, permitting experimental blinding. Obviously ill or suffering animals would have been excluded from the study, however no animal had to be excluded or died during the study.
All testing took place during the light phase of the daily cycle. All experiments were approved by the Institutional Animal Care and Use Committee (Project of Experiments No. 66/2016) and complied with the Animal Protection Act of the Czech Republic, EU Directive (2010/63/EC). The study was not preregistered.

Timing of procedures and group design
Ultrasonic vocalizations as a measure of mother-seeking behavior were recorded at P7 in a proportion of animals (for detailed description, see Supplementary Methods). At P12, DSE or control treatment was administered. All animals were attributed to four experimental groups depending on genotype and DSE status: wild-type naïve, Tsc2+/naïve, wild-type DSE and Tsc2+/-DSE. At P19, a proportion of juvenile rats were behaviorally assessed as described below. At the age of three months, the main experimental setup in adult rats was initiated, comprising assessment of behavior, epileptiform activity and everolimus concentration in the brain. The experimental schedule consisted of three time points. First, rats underwent experiments under baseline (non-medicated) condition (T1), followed by two weeks of treatment with everolimus and a second round of experiments under medicated condition (T2). After 8 weeks of washing-out, a third round of experiments was conducted (T3). A graphical summary of the order of procedures and investigations can be found in Fig. 2a and Table 1.

Pilocarpine-induced developmental status epilepticus in pups
To simulate early developmental epilepsy, we used pharmacologically-induced DSE. We decided to use lithium-pilocarpine-induced DSE at P12 instead of the previous paradigm of repeated kainic acid-induced DSE at P7 and P14 [32,33]. The models are epileptologically very similar, but in a pilot experiment lithium-pilocarpine exhibited stronger phenotype without mortality previously reported for kainic acid [32] ( Fig. 1, Suppl. Fig. S1). The lithium-pilocarpine seizure model is well characterized and has been successfully used as a model of early epilepsy [36,37].
Pilocarpine induction of DSE was performed following the protocol described by [38]. LiCl (127 mg/kg, i.p., dissolved in distilled water, purchased from Sigma Aldrich, Czech Republic, cat. no. L4408-100G) was applied to all pups at P11, to sensitize them to pilocarpine and to facilitate seizures [39]. After 24 hours (P12), half of the pups received a single i.p. injection of pilocarpine (35 mg/kg, injection volume 10 ml/kg, purchased from Sigma Aldrich, Czech Republic, cat. no. P6503-5G) and the others received 0.9% saline vehicle solution (naïve control). The state of the pups was checked by visual observation, heating pad was used to prevent hypothermia, and additional saline was applied after the DSE procedure to prevent dehydration. With regard to brain development, the time span between P7 to P14 in rats is comparable with the rst year of life in humans [40]. DSE in this period leads to life-long impairments of cognition and synaptic plasticity [41,42] and has already been described as a model of infantile spasms in TSC patients [33,40].
Behavioral testing of juveniles At P19, juvenile rats (n = 34) from the main experimental group were tested for spontaneous exploration of unknown environment and social behavior. In the open eld test (OF), the juvenile was placed into the middle of a dimly illuminated (18 lux) Plexiglas box (40 × 35 × 21 cm) with fresh bedding and recorded for elevated plus maze apparatuses at P74-P78, to eliminate novelty effect from the subsequent testing. The behavioral testing battery conducted at T1, T2 and T3 covered several behavioral domains affected in TSC and ASD patients: social behavior and communication (social interaction, SI), anxiety (elevated plus maze test, EPM; light-dark test, LD), spontaneous locomotor activity (open eld test, OF), learning and memory (Y-maze test; novel object recognition test, NORT). The design and temporal order of the tests is represented by Tables 1 and 2. The experimenter who handled the animal was always unaware of the experimental group assignment. Social interaction test (SI) In this test, two unfamiliar rats (non-cage mates) of the same age from the same group were interacting in the neutral, familiar environment of the open eld arena, dimly illuminated (18 lux). Social isolation (10 min) preceded the interaction test, which itself took 10 min, but only the rst 5 minutes were analyzed. The following parameters were evaluated: a) social behavior: anogenital exploration, non-anogenital exploration, climbing on or pinning the social partner, following or approaching the social partner, play/ ght, evade; b) non-social behavior: self-grooming, freezing.
Elevated plus maze test (EPM) The apparatus, made from light grey plastic, consisted of four arms (each 10 cm wide, 50 cm long), elevated 70 cm above the oor. Two arms opposite to each other were open, the other two were surrounded by 30 cm high walls. The apparatus was illuminated by uorescent tubes on the ceiling (480 lux in open arms, 85 lux in closed arms). Each animal was placed into the middle part facing into an open arm and recorded for 5 min. The animals were habituated to the apparatus by a 5min session one week before T1, to avoid the effect of novelty. Time spent in the open arms was evaluated as a classical measure of anxiety. Looking down from the maze (a form of risk assessment behavior also termed head dipping or scanning in the literature) was counted whenever the head of the rat protruded over the edge of an open arm. This behavior has been shown to provide a measure of anxiety, perhaps linked to decision-making [43,44]. The total number of arm visits was used as a measure of locomotor activity. Habituation to the arena was done one week before T1. Locomotor activity (total distance walked) was measured.
Novel object recognition test (NORT) Novel object recognition test was done in the same apparatus as OF, but with a dim, indirect illumination (18 lux). It was done 24 hours after the open eld test, to ensure familiarity of the animals with the environment. Two different pairs of identical objects of comparable size (glass jars full of pebbles and glass cuboid containers) were used. The objects were fastened by twosided tape to the oor in the opposite corners (20 cm from the walls) and cleaned with water after each session. The animal was put into the central part facing an empty corner. In the initial sampling session, the rat was left free in the arena to explore two identical objects for 5 min. After 15 min retention interval, one object was replaced by a novel one and the animal was allowed to explore them for another 5 min (discrimination session). The use of objects as familiar/novel was counterbalanced to minimize the effect of any eventual preference of object or place, unrelated to the experimental stimulus. Discrimination index, as a measure of ability to differentiate between familiar and unfamiliar objects, was calculated as (n-f)/(n + f); where 'n' is time spent by exploration of the novel object, while 'f' is the time spent exploring the familiar object.
Y-maze test Y-maze was used to study spontaneous alternation as a measure of working memory. The apparatus consisted of three identical arms (10 cm wide, 50 cm long, 30 cm high walls; labeled A, B and C) made from white plastic. The rat was placed into the central part facing the arm A and was left free to explore the maze for 8 min. Spontaneous alternation (in %) was calculated as number of correct triads of arm entries (ABC, BCA, CAB, ACB, CBA, BAC) divided by the number of all triads. A same arm entrance (AA, BB, CC) was counted when the rat came out of an arm to the center and returned back again to the same arm. Total arm visits were analyzed as an activity measure.

Everolimus administration
Everolimus suspension was provided by Novartis (Basel, Switzerland) in a proprietary vehicle at 20 mg/ml. For application, it was diluted in 0.9% saline and injected i.p. at a dose of 1 mg/kg (injection volume 1 ml/kg), as described previously [33]. The drug was administered every other day for 12 days (6 injections in total), and the testing battery under medication (T2) was initiated during the second week of administration. All animals were injected by the active substance, and within-subject design was used to assess the effects on behavior. The drug was always applied in the afternoon, after behavioral testing. Body weight of rats was measured every day before the injection. Everolimus levels in the brain were not affected by rat genotype or history of DSE, and remained stable in the period corresponding to the behavioral testing, as shown in Suppl. Fig. S3.

Everolimus levels in the brain tissue
To address the question of everolimus levels in the brain during the behavioral tests, we harvested the brains from the main experimental group immediately after the T3 behavioral testing (8-weeks after the end of everolimus therapy). The rats were decapitated under iso urane anaesthesia. Additional rats of the same origin, age and treatment (n = 16), were sacri ced at the end of the 14-day everolimus medication, corresponding to the end of T2.
To verify if everolimus concentrations during the T2 testing period were stable, we used a group of intact Wistar rats (RRID:RGD_13508588, Velaz Ltd., Czech Republic, n = 16). The rats undergoing the same everolimus treatment as the main experimental group were sacri ced either before the medication (baseline), or 24 h after the third or the sixth ( nal) everolimus injection, which corresponds to the beginning and the end of the T2 behavioral testing battery.
The rats were anaesthetized by iso urane and sacri ced by decapitation, their brains quickly removed, cooled by dry ice and stored at -80 °C until analysis. Approximately 100 mg of rat brain tissue was transferred to Eppendorf tube; 100 mg of bullets for homogenization and 1 ml of acetonitrile were added to each sample. The samples were thoroughly vortexed and homogenized using the Bullet Blender Gold

Stereotactic surgery
For EEG recordings, we used additional 18 male rats divided into the same groups: wild-type naïve (n = 5), Tsc2+/naïve (n = 3), wild-type DSE (n = 4) and Tsc2+/-DSE (n = 6). The animals were stereotactically implanted at P80 with 14 gold-plated epidural electrodes (Mill-Max Mfg. Corp., product number 310-93-132-41-001, purchased from Farnell Czech Republic), under general iso urane anesthesia (2.5%). Electrodes were implanted epidurally in homologous areas of the frontal, parietal and temporal regions of the right and left hemispheres. Coordinates were taken from Paxinos rat brain atlas, RRID:SCR_006369 [45]: A + 5 mm and L ± 2 mm for the frontal association cortex (electrodes F3/F4), A + 2.2 mm and L ± 3.2 mm for the primary motor cortex (electrodes C3/C4), A − 3.8 mm and L ± 2.5 mm (electrodes P3/P4) and A -4.5 mm and L ± 4.5 mm (electrodes P5/P6) for the lateral parietal association cortex, A -3.6 mm and L ± 7.2 mm for the temporal association cortex (electrodes T3/T4) and A -8.3 mm and L ± 5.8 mm for the secondary auditory cortex (electrodes T5/T6). Reference electrode was implanted above the olfactory bulb and ground electrode subcutaneously in the occipital region. Electrode positions are indicated in Fig. 4. Electrodes and the connector were xed to the skull with dental cement Dentalon (containing 1 g of active gentamicin per 100 g of Dentalon powder). After the surgery, rats were single-housed to prevent damage of the implant, and left for one week to recover.

EEG recordings
Four recordings were acquired from each animal. The rst one was done one week after implantation (Session 1 at P95 -age corresponding to T1 behavioral experiments). Then, everolimus was administered for two weeks prior to the second recording (Session 2 at P111, corresponding to T2 behavioral testing). Two more recordings were made during the washing-out period (Session 3 at P124 and Session 4 at P138), the latter corresponding to T3 behavioral testing.
Recording sessions were 42-55 minutes long (40 minutes of signal from each animal were analyzed), and were conducted in a box with bedding. The rats were able to move freely in the cage during EEG recording while connected to a data acquisition system.
Raw EEG signal was recorded using the BrainScope (M&I, Prague) BioSDA09 ampli er having a frequency band of 0.15-70 Hz. The system acquired data with a 16-bit depth, 7.63 nV/bit resolution (i.e. ∼130 bit/µV) and a dynamic range of ± 500 µV. The data were recorded using a sampling rate of 1000 Hz.

Data evaluation and statistical analysis
Sample sizes were adopted from previous behavioral experiments in the Tsc2+/-(Eker) rat [32,33]. In some substantially laborious experiments (EEG recordings), sample size was determined by the framework of experimental possibilities. With respect to our previous studies [32,33], anogenital exploration was the primary outcome. No animal was excluded from the study.
During all behavioral tests, the animals were recorded by an overhead video camera located above the apparatus. BORIS software [46] was used for o ine manual scoring of rat behavior; EthoVision software (RRID:SCR_000441, Noldus Information Technology, Wageningen, Netherlands) for automatic analysis of trajectory in OF. Ultrasonic calls were marked manually in the Audacity software (RRID:SCR_007198). EEG recordings were analyzed manually in BrainVision Analyzer 2.1 (RRID:SCR_002356), all ndings were video con rmed with behavior. All o ine analyses were blinded.
Data were statistically evaluated using the IBM SPSS Statistics 25 (RRID:SCR_019096). Three-way analysis of variance (ANOVA) with repeated measures was used for the main behavioral testing battery. MANOVA, t-test and one-way ANOVA were used when applicable. The factors included in the analysis were genotype (Tsc2+/-) and DSE as between-subject measures and time as within-subject measure. The factor of time covered the course of everolimus medication (T1 -baseline before treatment; T2 -under everolimus medication; T3 -after washing out). In case there was a signi cant genotype-time or DSEtime interaction, we used simple effects to specify the nature of the interaction, and to identify the testing condition where the groups differed. The data was log transformed to meet parametric assumption in case of non-normal distribution (indicated by Shapiro-Wilk test of normality). When transformation did not lead to normalization of the data (number of open arm visits in the elevated plus maze), the negative binomial model with log link function was used. Signi cance was accepted at p ≤ 0.05. When the assumption of sphericity was not met in ANOVA with repeated measures, we used Greenhouse-Geisser correction of df and p-values.
Statistical analysis of the data was independently veri ed using GraphPad PRISM 5, RRID:SCR_002798 (two-way ANOVA and Bonferroni post-hoc tests), also used to indicate signi cance in the gures. Data are shown as bar graphs with mean and SEM. With regard to the quantity of data combined in the gures, additional showing of individual data points resulted in overloaded presentations. Instead, full output tables of statistical results obtained from both types of statistical software can be found in Suppl. Tab. S1.

Severity of pilocarpine-induced DSE is independent of pup genotype
Pilocarpine injection at P12 caused diarrhea and seizures accompanied by automatic scratching, drumming motions, tremors, limb extensions, postural problems and wild running, which is consistent with signs described in the literature [47]. The signs appeared in characteristic order during seizure onset, and the time of their rst occurrence was independent of pup genotype (Fig. 1a). Seizure severity was assessed by visual observation complemented by video recordings at two 15-min observation intervals, the rst starting immediately after injection, and the second starting 4 hours after the injection. The signs were scored using a modi cation of the Racine scale. Clonic behaviors, such as shivering or forelimb clonus ("drumming ") counted as stage 3, while tonic behaviors (general stiffness, tail clonus, falls or failure of righting re ex) counted as stage 5. Behaviors which could represent both seizure-related automatisms and normal behavior (i. e. scratching) only counted as stage 1. The highest stage observed in the 15-min interval for each animal was then counted. Seizure severity was comparable in Tsc2+/and wild type pups at both time points (Fig. 1b). This suggests that genotype did not play a role in the severity of the seizure. After 5-6 hours, the symptoms receded in most individuals, and the pups were then returned to the mothers. No mortality was observed.

Social behavior is altered by Tsc2+/-and pilocarpineinduced DSE during early development
We monitored body weight of the animals from DSE induction to adulthood (P11-P90). We found an (p < 0.001); P19 (p = 0.001); P27 (p = 0.015) and P40 (p = 0.010), with DSE animals exhibiting lower body weight. This indicates that DSE affected physical development of the pups, with persistent retardation of growth only normalizing at early adulthood (Fig. 1c, 1d). Moreover, the effects of DSE on social behavior at P19 were examined. Total duration of social play was selected as the parameter most relevant to eventual autism-like de cits. The data were Log-transformed because of non-normal distribution, and analyzed by t-test. In pairs consisting of DSE pups, social play was much less common: T(15) = 3.15, p = 0.007. Also, ultrasonic vocalizations (total call count) were much less common in DSE pups: T(15) = 3.035, p = 0.008 ( Fig. 1e-g). Together, this con rms that DSE has a lasting impact on both physical and social development of the pups.
In experimentally naïve P7 pups (prior to DSE), we observed an altered isolation-induced ultrasonic vocalization (USV) pro le in Tsc2+/pups, suggesting that Tsc2+/also leads to early changes in vocal communication patterns (for detailed description, see Supplementary results, Fig. S2 and Suppl. Tab. S1).
In the social interaction test, the parameters of anogenital and non-anogenital social exploration were chosen for analysis (Fig. 2b, 2c (Fig. 2b). This indicates an effect of Everolimus on anogenital social exploration impaired by Tsc2+/-, but not by DSE. In non-anogenital social exploration, three-way ANOVA of T1-T3 was only signi cant for the effect of DSE F(1, 32) = 12.285; p = 0.001; p 2 = 0.277. Two-way ANOVA of T1, T2 and T3 showed an effect only of DSE (Fig. 2c). Other parameters either did not differ between groups, or were too rare for analysis, and are not shown.
Locomotor activity in the EPM, measured by total arm visits, was affected neither by Tsc2+/nor by DSE.
In three-way ANOVA, the effect of time was signi cant, F(3,96) = 3.839; p = 0.012; p 2 = 0.107, corresponding to a gradual decrease of activity across sessions, probably driven by habituation of the animals. (Fig. 3a). Open arm visits exhibited strongly non-normal distribution, which was not improved by data transformation. Therefore, we used Negative binomial model with log link function, and did not of Tsc2+/were present at habituation and T3, but not at T1 and T2 (Fig. 3c).
Locomotion of adult rats in the OF was neither affected by Tsc2+/nor DSE. In three-way ANOVA, there was an effect of time (F(3, 87) = 18.233; p < 0.001; p 2 = 0.386), again suggesting a gradual decrease of locomotion either due to habituation or age-dependent changes in activity (Fig. 3d).
In LD, the time spent in the light compartment, exhibited non-normal distribution and the data had to be didn't nd a speci c signi cant difference on any testing day, only a trend at T3 (p = 0.071). Two-way ANOVA for individual sessions did not nd speci c differences at T1 and T2, however, animals that underwent DSE did spend less time in the light compartment at T3 (Fig. 3e). Bonferroni's post-hoc tests showed that it was due to the Tsc2+/-DSE group exploring the light compartment signi cantly less than any other group (Supplementary Tab. S1).
In the Y-maze, locomotor activity indicated by the total number of arm visits, was only affected by time, F(2, 64) = 23.895; p < 0.001; p 2 = 0.427, with gradual decrease evidencing habituation to the environment.
Importantly, we veri ed that everolimus concentration in brain tissue at T2 was not signi cantly affected by either DSE or Tsc2+/- (Fig. 2d). It was also stable during the T2 behavioral testing period, and nondetectable after washing-out at T3 (Fig. 2d; Suppl. Fig. S3). Everolimus induced signi cant weight decrement, which was, however, quickly compensated after therapy discontinuation (Fig. 2e).
Tsc2 +/-and DSE induce epileptiform activity in adult rats which is not eliminated by everolimus EEG in combination with video monitoring was recorded during T1 (session 1), T2 (session2), four weeks later (session 3) and during T3 (session 4). EEG records showed normal and symmetric background activity and no signs of any generalized tonic-clonic seizures. However, epileptiform activity patterns were apparent in the data (Fig. 4, Supplementary Tab. S1).
During all sessions, patterns with frequency of 15-20 Hz (mean duration 576 ± 11 ms) were identi ed in the EEG (Fig. 4d) 0.136. The number of WDS was initially high, rapidly decreasing during the course of the session. Twoway ANOVA showed a disseminated pattern of differences caused by Tsc2+/and DSE (Fig. 4b).

Discussion
Our study aimed to investigate whether early epileptic seizure has an additional effect on the ASD-like phenotype in an animal model with Tsc2 haploinsu ciency, and whether the Tsc2+/and seizure-induced behavioral de cits differ in their responsivity to mTORi treatment. We found that the ASD-like social impairment induced by Tsc2+/responded well to everolimus therapy, con rming our earlier work [31][32][33], while the behavioral consequences of DSE were not ameliorated by the drug.
We found that pilocarpine-induced DSE led to long-term disturbance of physical and social development, corresponding to changes documented in the literature [48]. We also noted qualitative changes in isolation-induced USV of 7-days-old pups bearing Tsc2+/mutation. Altered vocalization patterns have been previously noted as an early manifestation of the ASD-like phenotype in Tsc2+/mouse pups [34]. In adult rats, both DSE and Tsc2+/mutation altered social behavior. While DSE led to general and persistent impairment of social exploration, Tsc2+/speci cally decreased anogenital social exploration in the baseline session. Anogenital and non-anogenital exploration behaviors seem to be quite independent from each other, so it may not be surprising to nd the one to be more sensitive to genetic factors than the other [49]. As anogenital sni ng is important for individual recognition in rats [50], we consider it to be particularly sensitive to social de cits.
An ASD-like phenotype was previously reported in both epilepsy-naive Tsc2+/-(Eker) rats and wild-type rats after kainic acid-induced DSE [32,33]. In a genetically comparable Tsc2+/mouse model, decreased social interaction was found by Sato et al. [15], although an earlier study reported no social impairment [13]. Other genetic manipulations causing more severe Tsc2 de ciency, or dysfunction of the TSC1-TSC2 complex, lead to abnormal social behaviors in mice [51,52].
Importantly, two-week therapy with mTORi everolimus reversed the effect of Tsc2+/on social behavior, with the bene cial effect persisting at least two months after discontinuation of the therapy. However, no such treatment response was observed in DSE animals. Although everolimus is known to suppress seizure frequency [9], it apparently did not ameliorate the persisting consequences of early developmental seizures. Our observations strongly suggest that Tsc2+/and DSE may induce social impairments via different neurobiological mechanisms, with only the former being sensitive to mTOR inhibitors.
Regarding non-social behavioral paradigms, we saw for both Tsc2+/and DSE groups an increase in risk assessment behavior in the EPM, for DSE corresponding to previous ndings [53,54]. Elevated anxiety was previously reported in the Tsc2+/-(Eker) rat [32], the ndings in mouse models of TSC are contradictory [52,55]. Taken together, anxiety appeared subtly increased in both Tsc2+/and DSE groups, which is compatible with the literature.
Tsc1+/-and Tsc2+/mutations are known to induce epilepsy in humans, but so far, a spontaneous epilepsy phenotype has not been described in the Tsc2+/rat or any comparable pre-clinical mouse models. However, previous studies assessed only overt behavioral signs of seizures [13,31,56,57], which severely limited their sensitivity. Tschuluun et al. [58] measured EEG in intact and prenatally irradiated Tsc2+/rats and detected no seizures, but their sample size was very small (n = 2 for intact Tsc2+/rats). Our EEG recordings show two classes of features which could be classi ed as epileptiform: wet dog shakes (WDS), possibly corresponding to focal limbic seizures [59], and intervals of rhythmic activity accompanied with immobility, both of which were more common in Tsc2+/rats. WDS were, to a weaker extent than in Tsc2+/rats, also observed after induction of DSE. Although epileptic status is a known trigger of epileptogenesis, very young animals seem to be more resilient to epilepsy initiation, and DSE before P17 did not lead to recurring seizures in adult rats [60], which is consistent with our observations.
However, the occurrence of WDS decreased in time both between sessions and within a single session, especially in the rats carrying Tsc2+/mutation, in a manner resembling habituation. This may suggest stress during the recording procedure as a possible trigger of epileptiform activity in the pre-disposed Tsc2+/and DSE rats. Recently, Taylor et al. [61] have shown EEG signals in healthy rats mimicking absence seizures, which, however, can be terminated by external stimuli, and therefore shouldn't be considered true epileptiform activity. It is well possible that the rhythmic activity in our recordings was of similar nature, and it would be premature to classify it as epileptic without further veri cation. In addition, the limited number of animals in our EEG sample and rather short recording time limit the strength of our conclusions, but the observations are certainly interesting and deserve further examination. The present data do not show any effect of everolimus on EEG activity, but given the limitations by sample size and design, we must consider this conclusion tentative.
Despite the limitations, we showed that both Tsc2+/and DSE rats exhibit distinct EEG changes, adding to the previously described increased sensitivity to urothyl-induced seizures [57] and kindling epilepsy [31]. In clinical contexts, it remains unclear whether epilepsy in TSC patients is attributable to the Tsc2+/mutation, is secondary to cortical tubers acting as epileptogenic foci [5] or may be a combination of both. Epileptiform patterns seen in the Tsc2+/rat argue for the former, as the brains of Tsc2+/rats are virtually tumor-free until advanced age [62].
The molecular mechanisms mediating the role of mTOR overactivation and epilepsy in the pathogenesis of ASD are only very partially understood. Aberrant synaptic protein synthesis induced by mTOR and MAPK overactivation was proposed to represent one possible pathway leading to autistic phenotypes [63]. Epilepsy sets in motion a cascade of events that include gene expression, sprouting of bers, the establishment of new synaptic contacts and thus, persisting neuronal changes [64]. Our ndings suggest Tsc2+/and DSE to be pathophysiologically divergent, as only the former responded to mTORi treatment.
Returning to the clinical trial ndings of mTORi in relation to TAND, the present study shows that the social impairment in a pre-clinical model responds to mTORi in a manner depending on its etiology: the effect of Tsc2+/mutation is ameliorated by everolimus, while the effect of DSE is persistent -at least when treated for two weeks at adult age. If validated in other studies, our ndings may have implications for clinical practice. Given that a signi cant proportion of individuals with TSC and ASD have early-onset epilepsy, our ndings suggest that their ASD symptoms may only be partially responsive to mTORi. Our ndings emphasize the importance of early seizure suppression as a fundamental approach against the development of ASD in TSC and in general [2,65,66]. Individuals with TSC and ASD who do not have comorbid epilepsy, may represent a subgroup that could show more clear-cut bene t from mTORi treatment. Our study does not predict the effects of early developmental or long-term treatment with mTORi.
Finally, we acknowledge the limitations of any translational research and the challenges to recapitulate mutational and behavioral equivalence between human and rodent models (62). However, we used an established heterozygous animal model of TSC, and used established paradigms of epilepsy and social behaviors, as used in pre-clinical research [67]. Our DSE paradigm with a single period of prolonged seizures was only an approximation of the situation in epilepsy patients, who experience recurring seizures throughout their lives. Further, it is unclear how valid pilocarpine-induced DSE is as a model for infantile spasms in TSC. It is however still remarkable that social behavior impairments in the epilepsy model could not be ameliorated by mTORi. Given that our study examined a two-week treatment with everolimus in adult rats, it may not represent possible effects of early developmental and/or long-term treatment with mTORi. In addition, we did not assess the molecular pathways and mechanistic state of the mTOR signaling cascade in experimental groups and treatment stages directly. However, mTORi status in the brain was estimated by analyzing everolimus levels. In spite the potential limitations, the ndings presented here are novel and provide direct experimental evidence for differential contributions of molecular and seizure-related mechanistic pathways to ASD-like phenotypes in TSC.

Conclusion
We report here rst, behavioral manifestations occured early in development of Tsc2+/and DSE rats; second, both the presence of a Tsc2+/mutation and a history of DSE decreased anogenital social exploration in adult rats, but only the de cit induced by Tsc2+/was ameliorated by mTORi treatment; and third, we showed that both Tsc2+/and DSE caused persisting epileptiform activity not eliminated by mTORi treatment. These ndings suggest that both Tsc2+/and DSE induce abnormalities in social behavior, but through different mechanisms. We propose that our ndings may contribute to an explanation why mTORi treatment was so far not effective in TSC individuals with ASD where comorbid early-onset epilepsy is common.

Consent for publication
Not applicable.