Fluoxetine Can Cause Epileptogenesis and Aberrant Neurogenesis in Male Wild-Type Mice

Abstract Antidepressants in general, and fluoxetine in particular, increase adult hippocampal neurogenesis (AHN) in mice. Here we asked how the antidepressant fluoxetine affects behavior and AHN in a corticosterone model of depression. In three groups of adult male C57BL/6j mice, we administered either vehicle (VEH), corticosterone (CORT) treatment to induce a depression-like state, or corticosterone plus a standard dose of fluoxetine (CORT+FLX). Following treatment, mice performed the open field test, the novelty suppressed feeding (NSF) test, and the splash test. Neurogenesis was assessed by means of immunohistochemistry using BrdU and neuronal maturation markers. Unexpectedly, 42% of the CORT+FLX-treated mice exhibited severe weight loss, seizures, and sudden death. As expected, the CORT-treated group had altered behaviors compared to the VEH group, but the CORT+FLX mice that survived did not show any behavioral improvement compared to the CORT group. Antidepressants generally increase neurogenesis and here we also found that compared to CORT mice, CORT+FLX mice that survived had a significantly greater density of BrdU+, BrdU+DCX+, and BrdU+NeuN+ cells, suggesting increased neurogenesis. Moreover, the density of BrdU+NeuN+ cells was increased in an aberrant location, the hilus, of CORT+FLX mice, similar to previous studies describing aberrant neurogenesis following seizures. In conclusion, fluoxetine could induce considerable adverse effects in wild-type mice, including seizure-like activity. Fluoxetine-induced neurogenesis increases could be related to this activity; therefore, proneurogenic effects of fluoxetine and other antidepressants, especially in the absence of any behavioral therapeutic effects, should be interpreted with caution.


Introduction
Adult hippocampal neurogenesis (AHN) occurs in the subgranular zone of the dentate gyrus (DG) in most mammals that have been tested.Whereas most immunohistochemical experiments provide evidence for human hippocampal neurogenesis, more experimental approaches are needed to understand its extent and functional relevance.Neurogenic rates can be altered pharmacologically, by diet, exercise, and environmental enrichment [1,2], most of which are also known to affect mood.These studies have prompted an approach of pharmacologically augmenting neurogenesis to ameliorate memory and mood in humans suffering from depression, Alzheimer's, and other diseases [2].
Depression is responsible for massive burden of morbidity in the UK and worldwide.Patients can spend significant proportions of their lives struggling with faulty cognition, impaired memory, and fractured social interactions.Poor pattern separation is frequently associated with depression [3] and is a cognitive function reliant on adult neurogenesis.There are no perfect animal models for depression, but corticosterone administration is useful as it replicates the stress, hormonal responses, and cognitive impairment of depression.Extensive preclinical [4][5][6][7][8] and some clinical [9,10] evidence shows that most classes of currently prescribed antidepressants increase AHN.Multiple studies have shown proneurogenic effects of selective serotonin reuptake inhibitors (SSRIs) and fluoxetine both in rodents and in clinical populations [11][12][13][14][15][16][17][18].Moreover, several studies suggest that neurogenesis is required for antidepressants to exert their therapeutic effect in mice [19], although most animal studies show that this effect is limited to particular behavioral or neurobiological endophenotypes [20][21][22].Cortisol concentrations are typically high in depression and patients frequently exhibit abnormal negative feedback in the dexamethasone suppression test.Thus, when patients take fluoxetine, it is likely that it will be in the presence of high concentrations of cortisol.In this study, we administered corticosterone (murine cortisol), which is considered to be a model of depression [21,23,24].
AHN can also be stimulated by a number of aversive stimuli, such as seizures [25,26], electroconvulsive therapy [4,27], stroke [28,29], and traumatic brain injury [30,31].In these studies, injury-induced neurogenesis is sometimes described as aberrant, with pathological features such as ectopic newborn granule cell migration and aberrant dendritic tree growth into the hilus, as opposed to the granular and molecular cell layers of the DG.Importantly, some of these studies show that blocking or reducing neurogenesis improves postinsult functional outcomes.As such, one study showed that ablation of neurogenesis prior to the induction of pilocarpineinduced seizures prevented the development of chronic seizures [32].Another study showed that ablation of neurogenesis following stroke improves cognitive outcomes [33].Finally, early increases in neurogenesis after concussive injury promotes seizure susceptibility [34].Thus, it appears that increased AHN could be detrimental in certain situations in contrast to its generally beneficial effects on neuropathology and could reflect a response to an adverse influence as well as a therapeutic one.
In this study, we report that fluoxetine treatment at levels similar to those used clinically and in similar mouse paradigms in a corticosterone-induced model of depression can induce seizure-like episodes, severe weight loss, and aberrant neurogenesis with minimal therapeutic behavioral effect.We suggest that these data should raise caution vis-à-vis the role of neurogenesis in treatment for depression.

Mice
All experiments were conducted in accordance with the United Kingdom Animals Scientific Procedures Act (1986) under project license PPL 30/3096 and were approved by the Local Ethical Review for the University of Oxford.7-Week-old male C57BL/6j mice were obtained from Charles River Laboratories, UK. Mice were housed at the Biomedical Sciences Building SPF facility at the University of Oxford, in individually ventilated cages with wood chip bedding in groups of four siblings per cage.Mice were maintained on a 12 h light/dark cycle (7:00 am to 7:00 pm).Upon arrival, three sibling groups (12 mice per group) were randomly allocated to one of the three treatment groups: control group (VEH), corticosterone-treated group (CORT), and corticosterone-and fluoxetine-treated groups (CORT+FLX).Experimental procedures began 1 week after the animals' arrival at the facility.

Treatments
The CORT and CORT+FLX groups were administered corticosterone (35 μg/mL equivalent to 5 mg/kg/day) in drinking water throughout the whole experiment, according to previously published protocols [21,23,24].Corticosterone powder (C2505, Sigma-Aldrich, UK) was dissolved in 0.45% hydroxypropyl-βcyclodextrin (332607 Sigma-Aldrich, UK).The VEH group received 0.45% hydroxypropyl-β-cyclodextrin in drinking water during the entire experiment.On weeks 4-6 of corticosterone administration fluoxetine or vehicle was administered via oral gavage.Fluoxetine hydrochloride (F132, Sigma-Aldrich, UK) was dissolved in sterile saline (Aqupharm 1) and administered at a daily dose of 18 mg/kg, which closely replicates the concentrations of fluoxetine used in the clinic [35] and in animal studies [23].VEH and CORT groups received a similar volume (100 μL) of saline via oral gavage daily.BrdU was administered to all animals on week 6 of CORT treatment (BrdU powder; B5002, Sigma-Aldrich, UK) and was administered in drinking water at a concentration of 1 mg/mL with 1% sucrose to improve palatability.

Behavioral Testing
Behavioral testing began at week 9, 2 weeks after the last fluoxetine or vehicle gavage.The experimenter conducting behavioral testing was blinded to the treatment group allocation.All behavioral tests were performed during the light phase.On the day of the testing, mice were transferred to the behavioral unit and allowed to acclimatize for at least 30 min prior to the start of the test.For the open field test, a circular brightly lit 50 cm diameter arena was used.Mice were placed inside the arena facing the wall.Mouse movements were recorded and analyzed using the Etho-Vision software (Noldus) for 20 min.For the novelty-suppressed feeding (NSF) test, mice were food deprived for 18 h prior to test.A rectangular arena with fresh wood chip bedding was used for the test.The food pellet was fixed in the middle of the arena.Latency to start eating the food pellet was manually recorded.If a mouse did not start eating the pellet during the first 10 min in the arena, its latency was recorded as 600 s.For the splash test [20], mice were sprayed with a 10% sucrose solution and placed in a novel cage for 5 min.Grooming behavior was recorded using EthoVision software and manual scoring method.

Immunohistochemistry
Animals were overdosed with euthatal and transcardially perfused with cold PBS and 4% paraformaldehyde.Brains were removed and post fixed in 4% paraformaldehyde overnight and then cryoprotected in 30% sucrose in 0.1 M phosphate buffer for 3 days.Brains were frozen in dry ice and then sectioned into 30 μm coronal sections using a freezing microtome (Leica).Immunohistochemistry was then performed as previously described [36].Anti-BrdU (1:400, NB500-169, Novus Bio), anti-DCX (1:100, SC8066, Santa Cruz), and anti-NeuN (1:300, MAB377, Millipore, USA) primary antibodies were used for IHC.Alexa Fluor secondary antibodies were used (1:500, Life Technologies).Antigen retrieval for BrdU staining was performed with 1 M HCl at 37°C for 1 h.Images were taken on an automated EVOS FL Auto2 (Invitrogen) cell and slide imaging system and manually quantified using ImageJ.An average of 3 sections per hippocampal region per brain was used for analysis.To determine tissue volume for density calculation, the granular zone was delineated according to [37], and the volumes determined based on the surface area and the height of the Z-stack.

Statistics
Data were expressed as mean ± standard error of mean (SEM).Data analysis was performed using GraphPad Prism 7.05.A twoway ANOVA with Dunnett's multiple comparisons test was used to analyze statistical difference between group means.For weight data, a repeated measures two-way ANOVA was used.

Results
In this experiment, mice were treated with corticosterone to induce depression-like behavioral and physiological changes and were treated with fluoxetine to counteract them (Fig. 1a).However, on the third week of fluoxetine administration, 5 of 12 treated mice experienced episodes of behavior indicative of seizures associated with significant weight loss (above 20%) (2way repeated measures ANOVA: effect of time F(6, 192) = 57.52,p < 0.0001; time × treatment F(18, 192) = 1.64, p = 0.054; Fig. 1a, b).These mice had to be culled.Interestingly, the remaining seven mice displayed increased weight gain compared to the CORT group, potentially underlying a dichotomous response to the antidepressant treatment (time × treatment F(18, 243) = 10.9, p < 0.0001).
The remaining mice were subjected to behavioral testing after a 2-week period of drug washout.Interestingly, compared to corticosterone-alone treated mice, CORT+FLX-treated mice spent less time in the inner zones of the open field test (F(2, 27) = 4.31, p = 0.024; Fig. 2a), suggesting that they were more anxious in the open field test than the CORT group.Locomotor activity measured in the open field did not differ among groups (Fig. 2b).Corticosterone treatment increased the latency to start eating in the NSF test, suggesting induced anhedonia as expected (Fig. 2c).However, fluoxetine treatment did not reverse this effect (F(2, 20) = 8.24, p = 0.0024; Fig. 2c).A similar effect was observed with grooming behavior assessed by means of the splash test.The CORT group showed reduced grooming compared to VEH (F(2, 27) = 3.45, p = 0.046; Fig. 2d), but fluoxetine did not improve this.
After behavioral testing, levels of AHN were assessed by measuring the density of cells labeled with the S-phase marker BrdU, the immature neuronal marker DCX, and the mature neuronal marker NeuN (Fig. 3a).We examined both the dorsal and the ventral hippocampus as the former regulates spatial memory, while the latter is more implicated in mood regulation [8].Immunohistochemical analysis showed that fluoxetine significantly increased the number of proliferating BrdU+ cells in the dorsal and ventral parts of the hippocampal DG (granular and subgranular zones) (treatment effect F(2, 23) = 27.3,p < 0.0001; Fig. 3b).In the dorsal DG, the CORT+FLX group had significantly more DCX+BrdU+ cells and NeuN+BrdU+ cells than the CORT alone group.Both CORT and CORT+FLX treatments had a significant effect on the level of DCX+BrdU+ double-positive cells in the ventral part of the DG, with CORT decreasing cell density compared to VEH and CORT+FLX restoring it to VEH levels (treatment effect F(2, 23) = 6.69, p = 0.0051; Fig. 3c).A similar effect was observed on the density of NeuN+BrdU+ double-positive cells (treatment effect F(2, 23) = 11.45,p = 0.0004; Fig. 3d).However, when we measured the percentage of DCX+BrdU+ cells among BrdU+ cells, this parameter was decreased in the ventral hippocampus in the CORT+FLX group (treatment effect F(2, 23) = 4.91, p = 0.017; Fig. 3e).No treatment modified the percentage of NeuN+BrdU+ cells among BrdU+ cells (Fig. 3f).To determine whether antidepressant treatment induced ectopic neurogenesis, we calculated the density of DCX+ and NeuN+BrdU+ cells in the hilus.The result showed that indeed FLX treatment increased the number of NeuN+BrdU+ cells between the blades of the DG, in the hilus, where neurogenesis is not expected to take place in adulthood (treatment effect F(2, 15) = 4.03, p = 0.04; Fig. 3g).

Discussion
We gave a standard dose of FLX, which closely replicates the concentrations of fluoxetine used in the clinic [35] and in animal studies [23].Nevertheless, we observed severe weight loss and seizures in 42% of the mice, which has not frequently been described in the literature.Occasional reports show that such effects of fluoxetine occur in some transgenic animals [38,39].Interestingly, clinical reports show that antidepressants have some epileptogenic activity, however, fluoxetine is reported to show the lowest level of risk [40,41].Moreover, the manufacturer of fluoxetine warns of seizure activity, especially in those already susceptible, on the drug label.Flx and Cort may have interacted, and their compound effects may have been synergistically toxic, resulting in seizures and death.BrdU was originally developed as a chemotherapeutic radiosensitizer and at high doses can be toxic [42].BrdU, fluoxetine, and corticosterone may have had systemic effects, for example, the latter affecting adrenal cortex physiology.However, it is important to note that the seizures and death occurred before BrdU was given.
No further data are available from the mice which suffered the severe side effects since the seizures were unexpected and they had to be culled.The other animals in the three groups CORT+FLX, CORT, and VEH did not display seizures and were analyzed for depression-like behavior and levels of AHN.Note that it is possible that some seizure activity or subictal states took place outside of the observation times in these animals.Interestingly, in contrast to mice with CORT+FLX-induced seizures, the animals that did not seize showed considerable weight gain once the antidepressant treatment was stopped.These data may be indicative of a bimodal response to fluoxetine and high levels of corticosterone.We do not understand the substrate for these differences, but similar results have been noted by others.Indeed, groups of C57BL6 mice were shown to diversify into treatmentresistant and treatment-respondent phenotypes upon fluoxetine treatment [43].However, our behavioral analysis of the CORT+FLX intact mice showed that fluoxetine did not exert any therapeutic effect on the depression-like behavior induced by chronic corticosterone treatment.This is consistent with clinical literature increasingly questioning whether SSRIs exert worthwhile antidepressant effects, reflected in recent guidance in England stepping away from the centrality of antidepressants in treating depression [44][45][46].
Corticosterone diminishes neurogenesis and our data also showed this effect as significantly fewer DCX+BrdU+ cells and NeuN+BrdU+ cells found in the ventral DG after CORT treatment.We chose to use 7-week-old mice as these are considered to be young adults, the period of life when depression frequently starts.Adult neurogenesis decreases significantly in older animals and humans with Alzheimer's disease [47,48].Therefore, it may be that in elderly populations, cortisol effects on decreasing neurogenesis would be even greater than predicted by this study.The ventral DG is a subdivision proposed to regulate mood and anxiety [49] and two of the key features of depression are decreased mood and increased anxiety.Thus, we expected these mice to also have behavioral changes reminiscent of depression.CORT administration augmented the latency to feed in the novelty-feeding suppression test, a behavior associated with anhedonia.CORT also reduced the time mice spent grooming themselves in the splash test, which may model anhedonia, or the diminished self-regard and reduced hygiene associated with depression.However, the behavioral tests of anxiety and locomotion were unchanged by CORT administration.Fluoxetine has been shown to reverse some of the behavioral defects associated with CORT administration; however, in our study, it only influenced one out of the four behaviors measured.In the novelty-suppressed feeding experiment, there was no statistically significant difference between the CORT and CORT+FLX groups, but there was a trend for fluoxetine to reduce the latency to start eating.A larger cohort of mice may reveal such an effect.It is also important to note  that the behavioral tests were conducted after a 2-week washout as opposed to during the fluoxetine treatment, which may have obscured the effect.In the CORT+FLX group, the amount of time spent in the inner zone of the open field test was reduced compared to the CORT mice.This was opposite to our prediction as the inner zone of the open field is anxiogenic in mice, but fluoxetine is thought to be anxiolytic.On the other hand, anxiogenic effects of fluoxetine have also been observed in the clinic [50].At the same time, it is possible that the CORT group was not less anxious but more anxious than the vehicle group and that FLX decreased that anxiety, thus removing freezing behavior.Overall, our results suggest that fluoxetine may not be as efficient in reducing depressionlike behaviors as was expected based on the literature.
The lack of behavioral response was not entirely reflected in the neurogenesis data.As expected based on the literature [4][5][6][7][8], fluoxetine induced a marked increase in the number of proliferating (BrdU+) cells, of newborn young neurons (DCX+BrdU+) and newborn maturing neurons (NeuN+BrdU+) in the DG.In fact, fluoxetine was able to rescue corticosterone-induced reduction in the density of DCX+BrdU+ as well as NeuN+BrdU+ cells in the ventral DG.Taken together, our work suggests that increasing neurogenesis in the CORT model of depression via fluoxetine can increase proliferation and restore neurogenesis to control levels but that this is not always correlated with behavioral rescue in the paradigms used (open field test, novelty supressed feeding, and splash test).It may be that with tests such as the conditioned fear response and Morris water maze, which more directly assess memory, correlations with neurogenesis would have been found.Unexpectedly, the percentage of BrdU+ cells that also expressed DCX was unchanged in the dorsal DG but decreased in the ventral DG of the CORT+FLX group compared to the VEH and CORT groups.Similarly, the percentage of NeuN+BrdU+/BrdU+ cells was unchanged in the dorsal and ventral DG.Since overall levels of BrdU were dramatically increased in the DG, this finding suggests that FLX may have preferentially induced the proliferation of other cell types, such as glia, as happens in response to brain tissue injury.
Additionally, we observed that in the CORT+FLX mice, increased BrdU+ cell density was not restricted to the granular and subgranular zones of the DG, but clearly extended into the hippocampal hilus and the molecular layer.Ectopic hilar neurogenesis has been reported in the context of antidepressant treatment and induced epilepsy [4].In our study, the density of adult-born NeuN+BrdU+ neurons was increased in the hilus of the CORT-FLX group mice.It may be that the hilar NeuN+BrdU+ cells induced by Cort/Flx were generated by subictal seizures [51,52].Thus, our findings may be explained by the CORT/FLX group experiencing nonfatal or subtle seizures sufficient to drive hilar neurogenesis.Overall, the potential glial proliferation and ectopic neurogenesis are suggested by our data points towards aberrant rather than normal healthy neurogenesis.It is possible that the CORT+FLX-intact mice also experienced seizures which were not detected by the experimenters or preictal activity which could have caused mild aberrant neurogenesis.Such a possibility raises the question of the frequency of fluoxetine's side effects in mice and the origins of the fluoxetine induction of neurogenesis in multiple previous studies.It also raises the wider possibility that neurogenesis detected in response to antidepressants in animal (and human) studies may represent a repair mechanism as is observed in the response to traumatic brain injury, stroke, and other noxious insults.Indeed, there is increasing attention paid to some of the toxic effects of antidepressants on the brain, manifested as prolonged withdrawal effects [53][54][55], post-SSRI sexual dysfunction [56], and cognitive deficits, and also implicated in cortical thinning [53,54].In conclusion, in this study we report that fluoxetine treatment in mice given corticosterone induced seizures and severe weight loss in a significant proportion of the animals.We also observed a lack of behavioral improvement after FLX and neurogenesis changes with features suggestive of aberrancy.Our results suggest that the neurogenesis increases reported upon fluoxetine treatment should be interpreted with caution, especially when behavioral data do not corroborate the therapeutic effects of the treatment.The efficiency of antidepressants still remains a focus of scientific debate [55], and increasingly their side effects profile, risk of dependence, and severity of withdrawal effects are becoming clear [57].The data presented here should inform clinical trials targeting AHN.Therefore, our findings make a noteworthy contribution to the research in the effects of antidepressant drugs.

Statement of Ethics
We confirm that we have read the Journal's position on issues involved in ethical publication and affirm that this report is consistent with those guidelines.The experiments were performed in accordance with the UK Animals (Scientific Procedures) 1986 Act, UK Home Office.All animal work was approved by the UK Home Office, License #30/2496, and the University of Oxford Department of Physiology, Anatomy and Genetics Departmental Ethical Review Committee.

Fig. 1 .
Fig. 1.Experimental design, survival, and weight changes during the experiment.a Experimental design.Mice (n = 12/group) were treated with CORT or VEH in drinking water throughout the experiment, and with FLX or VEH via gavage for 3 weeks starting from week 4 of experiment.Gavage was followed by 2 weeks of washout period and then by a period of behavioral testing.b Survival curves for treatment groups show the loss of animals in CORT+FLX group in the 3rd week of gavage treatment.c Weekly average weight measurements of the treatment groups analyzed using 2-way RP ANOVA.*p < 0.05, **<0.001derived from post hoc Dunnett's multiple comparisons CORT versus VEH or FLX groups.

3 (
For legend see next page.)

Fig. 3 .
Fig. 3. Neurogenesis measurements in experimental groups.a Triple immunohistochemistry in the three groups for NeuN, DCX, and BrdU.Note the increased number of BrdU+ cells in the hilus of the CORT+FLX group.The location of the molecular layer (mL), the dentate gyrus (dg) and the hilus (hl) are indicated in the NeuN VEH panel.b-g Show measurements in the dentate gyrus and G shows results in the hilus.b The density of proliferating BrdU+ cells was significantly increased in the CORT+FLX group compared to CORT alone in the dorsal and the ventral DG.c, d The density of adult-born DCX+BrdU+ and NeuN+BrdU+ neuroblasts was increased in the dorsal hippocampi of the CORT+FLX group compared to the CORT group.In the ventral DG, CORT alone decreased DCX+BrdU+and NeuN+BrdU+ density and FLX treatment was able to reverse this effect.e The percentage of DCX+BrdU+ cells among BrdU+ cells was not different in the dorsal hippocampus of the three groups.However, in the ventral hippocampus, the proportion of these cells was lower in the CORT+FLX group than in the CORT group.f The percentage of NeuN+BrdU+ cells among BrdU+ cells was not different in the three groups of mice.g The density of DCX+ neuroblasts was not significantly different in the hilus among the three groups.However, the density of NeuN+BrdU + cells was significantly increased in the hilus of CORT+FLX mice compared to the VEH group; *p < 0.05; **p < 0.01, ***p < 0.001 derived from Dunnett's multiple comparisons post two-way ANOVA, mean ± SEM.