Blockade of Serotonin 5-HT6 Receptor Constitutive Activity Alleviates Cognitive Deficits in a Preclinical Model of Neurofibromatosis Type 1

Neurofibromatosis type 1 (NF1) is a common inherited disorder caused by mutations of the NF1 gene that encodes the Ras-GTPase activating protein neurofibromin, leading to overactivation of Ras-dependent signaling pathways such as the mTOR pathway. It is often characterized by a broad range of cognitive symptoms that are currently untreated. The serotonin 5-HT6 receptor is a potentially relevant target in view of its ability to associate with neurofibromin and to engage the mTOR pathway to compromise cognition in several cognitive impairment paradigms. Here, we show that constitutively active 5-HT6 receptors contribute to increased mTOR activity in the brain of Nf1+/− mice, a preclinical model recapitulating some behavioral alterations of NF1. Correspondingly, peripheral administration of SB258585, a 5-HT6 receptor inverse agonist, or rapamycin, abolished deficits in long-term social and associative memories in Nf1+/− mice, whereas administration of CPPQ, a neutral antagonist, did not produce cognitive improvement. These results show a key influence of mTOR activation by constitutively active 5-HT6 receptors in NF1 cognitive symptoms. They provide a proof of concept that 5-HT6 receptor inverse agonists already in clinical development as symptomatic treatments to reduce cognitive decline in dementia and psychoses, might be repurposed as therapies alleviating cognitive deficits in NF1 patients.


Introduction
Neurofibromatosis type 1 (NF1) is a dominant autosomal disease with an estimated prevalence of about 1 in 3000 that is independent of ethnicity, race, or gender, and with a full penetrance. The hallmarks of the disease are "café au lait" spots and tumors of central and peripheral nervous systems, including neurofibromas, gliomas and pheochromocytomas [1]. Cognitive deficits represent another major feature of the disease, with up to 80% of the children with NF1 at risk of moderate to severe cognitive impairments that affect one or more areas of cognitive functioning and seriously compromise their scholar performance and quality of life [2,3]. These include learning disability, decreased attention, difficulties in executive planning and deficits in perception skills [4][5][6]. Moreover, at least half of children and adults with NF1 show social behavior impairments [7][8][9][10]. To date, clinical trials assessing the effect of statins or methylphenidate upon behavioral and cognitive outcomes yielded mitigated or poorly reproducible results [11][12][13][14][15], underscoring the need of new therapeutic strategies. , the data illustrated are the mean ± SEM of results obtained in three mice for the WT genotype and five mice for the Nf1 +/− genotype. * p < 0.05, unpaired t-test. (C). Western blots assessing p70S6K phosphorylation at Thr 421 -Ser 424 in PFC of Nf1 +/− mice injected with either vehicle (n = 3) or Rapamycin (Rapa, 10 mg/kg, n = 3), or SB258585 (SB, 2.5 mg/kg, n = 3) or CPPQ (2.5 mg/kg, n = 3) 30 min before sacrifice. Data are expressed as in B. * p < 0.05, one-way ANOVA followed by Dunnett's test. n.s.: not significant.

Effect of Blocking the 5-HT6 Receptor-mTOR Pathway on Sociability and Short-Term and Long-Term Social Memories of Nf1 +/− Mice
We first investigated sociability of Nf1 +/− mice by using the three-chamber social preference test (Figures 2A,D,G). After 10 min of habituation in the empty testing arena, mice were able to interact either with an inanimate object or an unfamiliar congener (sociability phase). Nf1 +/− mice spent more time interacting with their congener than with the object (p . Representative Western blots assessing p70S6K phosphorylation at Thr 421 -Ser 424 as an index of mTOR activity in PFC of adult WT and Nf1 +/− mice. Data represent the ratios of immunoreactive signals of the anti-phospho-Thr 421 -Ser 424 -p70S6K antibody to the immunoreactive signal of the anti-p70S6K antibody and are expressed in % of values in WT mice. In (A,B), the data illustrated are the mean ± SEM of results obtained in three mice for the WT genotype and five mice for the Nf1 +/− genotype. * p < 0.05, unpaired t-test. (C). Western blots assessing p70S6K phosphorylation at Thr 421 -Ser 424 in PFC of Nf1 +/− mice injected with either vehicle (n = 3) or Rapamycin (Rapa, 10 mg/kg, n = 3), or SB258585 (SB, 2.5 mg/kg, n = 3) or CPPQ (2.5 mg/kg, n = 3) 30 min before sacrifice. Data are expressed as in B. * p < 0.05, one-way ANOVA followed by Dunnett's test. n.s.: not significant. Then, we assessed preference for social novelty by performing the social discrimination test. After a 5-min retention interval, a novel congener was presented to the tested mouse, in addition to the familiar one ( Figure 2D). Both vehicle-treated WT and Nf1 +/− mice spent significantly more time interacting with the novel mouse, compared to the familiar one (p ˂ 0.001, n = 16 and p ˂ 0.001, n = 18, for vehicle-treated WT and Nf1 +/− mice, respectively, two-way ANOVA followed by Bonferroni's test, Figure 2E and Table 1). Both vehicle-treated WT and Nf1 +/− mice displayed a similar discrimination index (p > 0.05, Kruskal-Wallis followed by Dunn's test, Figure 2F and Table 1). However, after a 24-h retention interval, social novelty discrimination was impaired in Nf1 +/− mice, compared to WT mice (p < 0.001, Kruskal-Wallis followed by Dunn's test, Figures 2H,I and Table 1). We then explored whether blocking 5-HT6 receptor-operated mTOR signaling reverses long-term deficit in social discrimination observed in Nf1 +/− mice. After a 24-h retention interval, Nf1 +/− mice injected with SB258585 or rapamycin spent significantly more time interacting with the novel mouse, compared to the familiar one, in the social discrimination test (p ˂ 0.001, n = 20 and p ˂ 0.001, n = 12, for SB258585-and rapamycin-treated Nf1 +/− mice, respectively, two-way ANOVA followed by Bonferroni's test, Figure 2H,I and (D). For assessing short-term social discrimination (5 min after the sociability phase), the lateral compartments of the three-chamber apparatus contain a wire cage with either the familiar mouse or a novel mouse (E). Exploration time (expressed in %) of the novel and the familiar mouse by the tested mice. *** p < 0.001, significantly different from familiar mouse, two-way ANOVA followed by Bonferroni's test, with novelty and treatment as factors. (F). Discrimination index in each condition (WT vehicle: n = 16, Nf1 +/− vehicle: n = 18, Nf1 +/− SB: n = 23, Nf1 +/− Rapa: n = 12, Nf1 +/− CPPQ: n = 15). (G). For assessing long-term social discrimination phase (24 h following the short-term social discrimination test), the lateral compartments of the three-chamber apparatus contain a wire cage with either the familiar mouse or a novel mouse. (H). Exploration time (expressed in %) of the novel and the familiar mouse by the tested mice. n.s. not significant, *** p < 0.001, significantly different from familiar mouse; two-way ANOVA followed by Bonferroni's test, with novelty and treatment as factors. (I). Discrimination index in each condition (WT vehicle: n = 14, Nf1 +/− vehicle: n = 16, Nf1 +/− SB: n = 20, Nf1 +/− Rapa: n = 12, Nf1 +/− CPPQ: n = 15). n.s. not significant, * p < 0.05, ** p < 0.01, *** p < 0.001 vs. vehicle-treated Nf1 +/− mice, Kruskal-Wallis followed by Dunn's test.
Then, we assessed preference for social novelty by performing the social discrimination test. After a 5-min retention interval, a novel congener was presented to the tested mouse, in addition to the familiar one ( Figure 2D). Both vehicle-treated WT and Nf1 +/− mice spent significantly more time interacting with the novel mouse, compared to the familiar one (p < 0.001, n = 16 and p < 0.001, n = 18, for vehicle-treated WT and Nf1 +/− mice, respectively, two-way ANOVA followed by Bonferroni's test, Figure 2E and Table 1). Both vehicle-treated WT and Nf1 +/− mice displayed a similar discrimination index (p > 0.05, Kruskal-Wallis followed by Dunn's test, Figure 2F and Table 1). However, after a 24-h retention interval, social novelty discrimination was impaired in Nf1 +/− mice, compared to WT mice (p < 0.001, Kruskal-Wallis followed by Dunn's test, Figure 2H,I and Table 1). We then explored whether blocking 5-HT 6 receptor-operated mTOR signaling reverses long-term deficit in social discrimination observed in Nf1 +/− mice. After a 24-h retention interval, Nf1 +/− mice injected with SB258585 or rapamycin spent significantly more time interacting with the novel mouse, compared to the familiar one, in the social discrimination test (p < 0.001, n = 20 and p < 0.001, n = 12, for SB258585-and rapamycin-treated Nf1 +/− mice, respectively, two-way ANOVA followed by Bonferroni's test, Figure 2H,I and Table 1). However, Nf1 +/− mice that received an acute injection (i.p.) of CPPQ showed similarly altered performances to vehicle-treated Nf1 +/− mice (p > 0.05, Kruskal-Wallis followed by Dunn's test, Figure 2H,I and Table 1).
Neither SB258585 nor CPPQ nor rapamycin administration modified the sociability performance (p > 0.05, Kruskal-Wallis followed by Dunn's test, Figure 2B and Table 1) and short-term social memory of Nf1 +/− mice (p > 0.05, Kruskal-Wallis followed by Dunn's test, Figure 2C and Table 1). Likewise, SB258585, CPPQ or rapamycin administration to WT mice did not alter their sociability and their short-and long-term social memories (Supplementary Figure S2 and Table S1). 6 Receptor-Operated mTOR Signaling in Associative Memory Deficit of Nf1 +/− Mice Impairments in executive function and mental flexibility have been reported in patients with NF1 [35]. We used the object-in-place task, a behavioral task assessing associative memory and involving mental flexibility and executive function [36,37]. In this test, mice must distinguish between four familiar objects in their original (familiarization phase) or a novel configuration (test phase). Memory retention was evaluated 1 h after the familiarization session. Vehicle-treated Nf1 +/− mice did not discriminate between familiar and novel object configurations, as shown by the similar time they spent to explore the swapped object compared to the non-swapped objects during the test phase (p > 0.05, n = 11, two-way ANOVA followed by Bonferroni's test, Figure 3 and Table 1), whereas vehicle-treated WT mice discriminated between familiar and novel object configurations (n = 13, p < 0.001, Kruskal-Wallis followed by Dunn's test, Figure 3 and Table 1). Injection of SB258585 or rapamycin to Nf1 +/− mice restored ability to discriminate between familiar and novel object configurations (p < 0.05 for Nf1 +/− SB vs. Nf1 +/− vehicle and Nf1 +/− Rapa vs. Nf1 +/− vehicle, Kruskal-Wallis followed by Dunn's test, Figure 3 and Table 1). In contrast, CPPQ-injected Nf1 +/− mice were still not able to discriminate novelty in the object-in-place test (p > 0.05 for Nf1 +/− CPPQ vs. Nf1 +/− vehicle, Figure 3 and Table 1). Injection of rapamycin or SB258585 or CPPQ to WT mice did not affect their performance in the object-in-place task (Supplementary Figure S3 and Table S1). Table 1). However, Nf1 +/− mice that received an acute injection (i.p.) of CPPQ showed similarly altered performances to vehicle-treated Nf1 +/− mice (p > 0.05, Kruskal-Wallis followed by Dunn's test, Figures 2H,I and Table 1).

Role of 5-HT
Neither SB258585 nor CPPQ nor rapamycin administration modified the sociability performance (p > 0.05, Kruskal-Wallis followed by Dunn's test, Figure 2B and Table 1) and short-term social memory of Nf1 +/− mice (p > 0.05, Kruskal-Wallis followed by Dunn's test, Figure 2C and Table 1). Likewise, SB258585, CPPQ or rapamycin administration to WT mice did not alter their sociability and their short-and long-term social memories (Supplementary Figure S2 and Supplementary Table S1).

Role of 5-HT6 Receptor-Operated mTOR Signaling in Associative Memory Deficit of Nf1 +/− Mice
Impairments in executive function and mental flexibility have been reported in patients with NF1 [35]. We used the object-in-place task, a behavioral task assessing associative memory and involving mental flexibility and executive function [36,37]. In this test, mice must distinguish between four familiar objects in their original (familiarization phase) or a novel configuration (test phase). Memory retention was evaluated 1 h after the familiarization session. Vehicle-treated Nf1 +/− mice did not discriminate between familiar and novel object configurations, as shown by the similar time they spent to explore the swapped object compared to the non-swapped objects during the test phase (p > 0.05, n = 11, two-way ANOVA followed by Bonferroni's test, Figure 3 and Table 1), whereas vehicle-treated WT mice discriminated between familiar and novel object configurations (n = 13, p < 0.001, Kruskal-Wallis followed by Dunn's test, Figure 3 and Table 1). Injection of SB258585 or rapamycin to Nf1 +/− mice restored ability to discriminate between familiar and novel object configurations (p < 0.05 for Nf1 +/− SB vs. Nf1 +/− vehicle and Nf1 +/− Rapa vs. Nf1 +/− vehicle, Kruskal-Wallis followed by Dunn's test, Figure 3 and Table 1). In contrast, CPPQinjected Nf1 +/− mice were still not able to discriminate novelty in the object-in-place test (p > 0.05 for Nf1 +/− CPPQ vs. Nf1 +/− vehicle, Figure 3 and Table 1). Injection of rapamycin or SB258585 or CPPQ to WT mice did not affect their performance in the object-in-place task (Supplementary Figure S3 and Supplementary Table S1).

Discussion
Loss-of-function mutations in the NF1 gene encoding the Ras-GAP neurofibromin lead to Ras disinhibition and subsequent overactivation of Ras-dependent pathways, including the mTOR pathway. Consistently and corroborating previous findings [38][39][40], enhanced mTOR activity was observed concomitantly with reduced neurofibromin expression level in PFC of Nf1 +/− mice, compared with WT mice. Cerebral mTOR can be activated by a large variety of extracellular signals. These include growth factors such as brain-derived neurotrophic factor, insulin, insulin-like growth factor 1 and vascular endothelial growth factor, which activate mTOR via their cognate tyrosine kinase receptors, and guidance molecules, such as reelin. A number of neurotransmitters including glutamate, dopamine, serotonin and endocannabinoids, are also known to activate the mTOR pathway in neurons through the stimulation of ionotropic or G protein-coupled receptors [41]. Among the 14 serotonin receptor subtypes, the 5-HT 6 receptor physically interacts with and activates the mTOR complex 1 (mTORC1) in several brain regions upon agonist receptor stimulation. This receptor also physically interacts with neurofibromin, suggesting that neurofibromin might negatively regulate receptor-operated mTOR signaling and, conversely, that 5-HT 6 receptors might contribute to the enhanced mTOR activity observed in Nf1+/− mouse brain. Consistent with this hypothesis, this non-physiological mTOR activation was abrogated by the peripheral administration of the specific 5-HT 6 receptor antagonist SB258585, which thereby reproduced the effect of rapamycin treatment.
A recent study reported elevated levels of serotonin in whole brain of Nf1 +/− mice, compared to WT mice [42]. This elevation of cerebral serotonin levels might lead to persistent 5-HT 6 receptor stimulation and subsequent overactivation of the mTOR pathway. On the other hand, 5-HT 6 receptors are known to exhibit a high level of constitutive activity, not only in cell lines expressing high receptor densities, but also in the mouse brain [21]. Constitutive activity of 5-HT 6 receptor is critically dependent of its dynamic association with protein partners, including neurofibromin [21,43]. Furthermore, we previously demonstrated that SB258585 behaves as a 5-HT 6 receptor inverse agonist not only the toward canonical Gs signaling but also mTOR signaling [23]. Here, we show that in contrast to what was observed in SB258585-treated mice, administration of CPPQ, a well-characterized 5-HT 6 receptor neutral antagonist, does not affect mTOR activity in PFC of Nf1 +/− , suggesting that 5-HT 6 receptor constitutive activity rather than agonist receptor stimulation contributes to mTOR overactivation in Nf1 +/− mouse brain.
Several preclinical studies using genetically engineered NF1 mouse models or established human tumor cell lines have demonstrated that mTOR overactivation underlies tumor proliferation in NF1, underpinning the therapeutic potential of rapamycin and its analogs [44][45][46]. Further supporting that mTOR inhibition may represent a relevant therapy for brain tumors in NF1, preliminary clinical studies confirmed unequivocal efficacy of rapamycin and its analogs to reduce the growth of plexiform neurofibroma [47] or low-grade glioma [48] observed in some patients with NF1. A large body of evidence also indicates a deleterious influence of aberrant mTOR signaling in the central nervous system upon cognition, in addition to its tumor growth promoting effects. Overactivation of mTOR has been involved in cognitive deficits observed in preclinical models of neurodevelopmental disorders, such as tuberous sclerosis, Fragile X syndrome and schizophrenia [20,[49][50][51][52][53]. Likewise, non-physiological mTOR activation underlies cognitive deficits observed in preclinical models of acute cannabis consumption in adulthood and of cannabis abuse during adolescence [22,54] and those observed in neuropathic pain conditions [23,53,55]. Intriguingly, the role of mTOR overactivation in cognitive deficits associated with NF1 has so far not been investigated, while learning disabilities and deficits in attention, executive functions and perception skills are common complications seen in the majority of children carrying mutations in the NF1 gene. In the present study, we show that Nf1 +/− mice exhibit a normal sociability and short-term social memory, but alterations in long-term social memory and associative memory that are both abolished by an acute injection of rapamycin, indicating that deregulation of mTOR activity also plays a critical role in NF1-associated cognitive deficits. Notably, rapamycin did not affect social and associative memories in WT mice, indicating that only a non-physiological mTOR activation, such as that measured in PFC of Nf1 +/− mice, affects these cognitive processes.
Corroborating our results implicating constitutively active 5-HT 6 receptors in the enhanced mTOR activity in PFC of Nf1 +/− mice, we also demonstrated that the associated deficits of long-term social memory and associative memory are abolished by peripheral administration of the 5-HT 6 receptor inverse agonist SB258585, but not a neutral antagonist (CPPQ). These results thus extend to NF1 previous observations indicating a critical influence of mTOR signaling under the control of 5-HT 6 receptors, in cognitive deficits associated with pathological conditions of different etiologies, including preclinical models of schizophrenia or cannabis abuse during adolescence [20,22].
The mechanisms underlying the deleterious influence of 5-HT 6 receptor-dependent mTOR overactivation upon cognitive functions in NF1 remain to be elucidated. A previous study showed that the specific deletion of neurofibromin in GABAergic neurons, but not pyramidal neurons, results in deficits in spatial learning (Morris water maze) that are caused by enhanced Erk activation in GABAergic neurons, increase in GABA release and subsequent deficits in hippocampal long-term potentiation LTP [56]. Together with the present results, these findings suggest that both enhanced Erk and mTOR activities contribute to cognitive impairment in NF1 and that deregulation of each pathway affects distinct cognitive functions. As Erk, mTOR finely tunes synaptic plasticity mechanisms such as LTP and long-term depression [41]. It is likely that overactivation of mTOR in PFC, one of the major brain structures involved in the control of cognition by 5-HT 6 receptor, likewise leads to perturbations of synaptic transmission and synaptic plasticity in this brain region. Another study revealed the importance of deregulation of hyperpolarizationactivated cyclic-nucleotide-gated channel 1 (HCN1, a neurofibromin-interacting protein) in the pathophysiology of NF1-associated cognitive deficits [57]. HCN1 is the predominant isoform of HCN channels, a family of voltage-gated channels that modulates neuronal excitability [58,59]. The contribution of HCN1 to perturbed cognition in NF1 is reminiscent of our observations in a mouse model of cannabis abuse during adolescence indicating that this treatment increases HCN1 activity and subsequently affects excitatory-inhibitory balance through a mechanism dependent of mTOR activation by 5-HT 6 receptors [22]. Whether mTOR under the control of 5-HT 6 receptors likewise affects HCN1 activity and excitatory-inhibitory balance in Nf1 +/− mice remains to be explored.
In conclusion, the present study provides a proof of principle that 5-HT 6 receptor inverse agonists already in clinical development as symptomatic treatments to reduce cognitive decline in dementia such as Alzheimer's disease or psychoses such as schizophrenia, might be repurposed as first-line treatments to alleviate alterations of cognitive functions in NF1 patients before the appearance of malignancies. Such a strategy would certainly be more relevant than direct mTOR blockade by pharmacological inhibitors at early disease stages, as 5-HT 6 receptor inverse agonists are well tolerated and will thus not reproduce the severe side effects induced by mTOR inhibitors and related to their immunosuppressant actions. Given the influence of mTOR in the growth of benign and malignant cerebral tumors associated with NF1 and the role of constitutively active 5-HT 6 receptors in the enhanced cerebral mTOR activity in Nf1 +/− mice, the impact of 5-HT 6 receptor inverse agonists on NF1-associated tumors certainly warrants further exploration.

Animals
Nf1 +/− mice (Brannan et al., 1994) were F1 hybrids from a cross between C57BL/6J and 129S2/SvPasCrl mice (Charles River France, L'Arbresle, France). Two-month-old mice from both sexes were used. All experiments used wild type littermates as controls. Mice were housed under standardized conditions with a 12-h light/dark cycle, stable temperature (22 ± 1 • C), controlled humidity (55 ± 10%) and free access to food and water. Animal husbandry and experimental procedures were performed in compliance with the animal use and care guidelines of the University of Montpellier, the French Agriculture Ministry and the European Council Directive (86/609/EEC).

Behavioral Tests
The sample size was assessed using the G power software. It was set with the mean value of 12, based on our previous results assuming a significance level of 5% and a power of 80%.

Social Behavior Tests
Testing was carried out in a rectangular, three-chamber box with dividing walls made of clear Plexiglas and an open middle section, which allows free access to each chamber [60]. The social interaction test was conducted as previously described [30]. Briefly, mice were extensively handled one week before behavioral analysis. Over two days, four 10-min sessions were conducted. First, a habituation session was conducted in which the tested mice were placed for 10 min in the middle compartment of the threechamber device without the walls between the compartments, to allow free access to the three compartments and the empty cylindric containment cages placed in the lateral compartments. After a 5-min retention interval, a social interaction session was conducted for 10 min (mouse of same age and sex as tested mice or inanimate object placed in the containment cages). Five min after the social interaction session, a short-term social discrimination test was conducted for 10 min (familiar mouse and a novel mouse of same age and sex as tested mice, in the containment cages). Following the short-term social discrimination session, the «novel» mouse was removed from the apparatus, and the tested mouse was allowed to interact with the «familiar» mouse for an additional 45 min. A long-term social discrimination session (familiar mouse and a novel mouse of same age and sex as tested mice, in the containment cages) was carried out 24 h after the short-term social discrimination session. Duration of direct contacts between the tested mouse and the cage housing the conspecific or the object was recorded for 3 min. The experiments were video-recorded and exploration times (nose in contact or sniffing at <1 cm) were measured by a blinded observer. Sociability (exploration time of the mouse-exploration time of the object/total exploration time) and discrimination indexes (exploration time of the novel mouse-exploration time of the familiar mouse/total exploration time) and percentages of exploration were compared between groups. Drugs were administered 15 min before the habituation session on day 1 and 45 min before the long-term social discrimination on day 2. Animals that did not receive drugs were injected with an equivalent volume of vehicle. Animals that stayed only in the central chamber during one of the 3 steps of the test were excluded.

Object-in-Place Task
Testing was carried out in an arena (50-cm width, 50-cm length, 50-cm height) placed in a dimly lit room with clearly visible contextual cues (black on white patterns) on the surrounding walls. Mice were habituated to the arena on day 1 for 10 min. On day 2, mice had a 10-min familiarization session with four different objects presented and placed in the four corners of the arena. Mice were transferred back to the home cage during 1-h retention interval before a 3-min training session with two objects swapped. The objects were plastic toys (3-cm width, 3-cm length, 5-cm height) and were cleaned with 20% ethanol between sessions. The experiments were video-recorded and exploration times (nose in contact with or sniffing objects at <1 cm) were measured by a blinded observer. Discrimination indexes (exploration time of novel object-exploration time of familiar object)/total object exploration time) were compared between groups. Mice with total exploration time of less than 3 s in the test session were excluded.