Dexmedetomidine does not compromise neuronal viability, synaptic connectivity, learning and memory in a rodent model

Recent animal studies have drawn concerns regarding most commonly used anesthetics and their long-term cytotoxic effects, specifically on the nervous tissue. It is therefore imperative that the search continues for agents that are non-toxic at both the cellular and behavioural level. One such agent appears to be dexmedetomidine (DEX) which has not only been found to be less neurotoxic but has also been shown to protect neurons from cytotoxicity induced by other anesthetic agents. However, DEX’s effects on the growth and synaptic connectivity at the individual neuronal level, and the underlying mechanisms have not yet been fully resolved. Here, we tested DEX for its impact on neuronal growth, synapse formation (in vitro) and learning and memory in a rodent model. Rat cortical neurons were exposed to a range of clinically relevant DEX concentrations (0.05–10 µM) and cellular viability, neurite outgrowth, synaptic assembly and mitochondrial morphology were assessed. We discovered that DEX did not affect neuronal viability when used below 10 µM, whereas significant cell death was noted at higher concentrations. Interestingly, in the presence of DEX, neurons exhibited more neurite branching, albeit with no differences in corresponding synaptic puncta formation. When rat pups were injected subcutaneously with DEX 25 µg/kg on postnatal day 7 and again on postnatal day 8, we discovered that this agent did not affect hippocampal-dependent memory in freely behaving animals. Our data demonstrates, for the first time, the non-neurotoxic nature of DEX both in vitro and in vivo in an animal model providing support for its utility as a safer anesthetic agent. Moreover, this study provides the first direct evidence that although DEX is growth permissive, causes mitochondrial fusion and reduces oxygen reactive species production, it does not affect the total number of synaptic connections between the cortical neurons in vitro.

Anesthetic agents are commonly used in patients undergoing various surgical procedures [1][2][3] and their safety and efficacy has improved significantly over time 4 . Notwithstanding the safety of the newer compounds, evidence is emerging to suggest that numerous clinically used anesthetics may cause cytotoxicity 5 -especially when tested on animal models. The evidence of harm in the context of humans however remains polemical 6 . This information is particularly important in the context of both developing and aging brains when neurons rely upon their activity and connectivity for communications and the formation of new synapses underlying learning and memory. Blocking activity or altering neuronal activity patterns, leads either to a reduction in synaptic strength or to the elimination of neuronal activity altogether 7 . Since most anesthetics exert their actions by suppressing neuronal activity and shutting off synaptic communications, it is believed that their neurotoxic effects likely involve both structures and functions that are deemed essential for neuronal excitability. Specifically, any compound that inhibits electrical activity during the critical time window encompassing synaptogenesis is likely to alter the network's assembly. For instance, blocking electrical activity during development severely compromises not only neuronal projections but also synaptic connectivity 8 . It therefore stands to reason that blocking neuronal activity with anesthetics, specifically when synaptic connections are being formed en masse may affect brain development and connectivity. However, it remains to be determined whether the anesthetic-induced perturbation of synaptic Dexmedetomidine treatment. Sprague-Dawley pregnant rats were purchased from a commercial breeder (Charles River Laboratories, Senneville, QC, Canada). Gestational progress was monitored from E16.5 until the date of birth (post-natal day zero-P0).
P7 rat pups of both sexes were injected with 25 µg/kg of dexmedetomidine subcutaneously either once (1×) or twice (2×), (24 h after the first injection), using a 0.1 ml volume. Control animals were injected with the same volume of saline (sodium chloride 0.9% B Braun, Mississauga, ON, Canada). The core body temperature of all animals was maintained with the use of a heated blanket and gloves filled with warm water replaced every 20 min. The respiratory status of the animal was monitored by counting breaths per minute monitored every 10 min following the first injection for the following 30 min, to ensure the pup's health. Furthermore, oxygen saturation (SpO 2 ) was monitored every 30 min after the first injection for the period of an hour to further ensure the pup's health and reception to the compound (Nonin Medical, Minneapolis, MN, USA).
Each pup's ear was notched at P14 and pups were weaned in cages of 2-3 animals, separated by sexes. Animals were kept undisturbed until P60 and then subjected to various behavioural tests.
Morris water maze. A 1.8 m diameter pool of water with a platform hidden 1 cm below the water surface (25 °C) was placed in a room that had remote visual cues on four different walls. A P60 rat was placed in the water first at the quadrant closest to the hidden platform (Q1) facing the pool walls and left to swim in the water until it reached the platform or not within an allocated 60 s period. If the animal had found the platform, it was left there for 15 s and then placed back into an empty cage with bedding for 45 s before placing it in the next 3 quadrants (Q2, Q3 and Q4). If the animal did not reach the platform before 60 s, the software tracked the latency as 60 s. This process was repeated for 5 consecutive days and latency (time taken) to reach the platform and distance swum were measured by the automated software, ANY-Maze (Stoelting Co, Wood Dale, IL, USA). Latency and distance were used as the learning and memory parameters characterized over the period of days 1 and 5.
Novel object recognition. Object selection was carried out according to established parameters that allow rats to discriminate between objects of similar complexity to avoid potential preference that might otherwise compromise the results 38 .
For 2 days, the P67 animals went through the first phase, habituation, which consisted of the placement of an animal in a square empty box of 60 cm per side with bedding on the floor and left to interact with the new environment for 15 min per day. On the third day, two identical objects were placed diagonally in the box and the animals were left to interact with these objects for 5 min during the familiarization phase. The animals were then taken from the box and placed in an empty cage for a 5 min retention interval. In the final step, testing was assessed for 3 min by replacing one of the familiar objects with a novel one and the ratio of the time spent interacting with the new object was monitored and measured to inquire on recognition learning and memory. To avoid bias, the object replaced to be the novel object changed for each animal (e.g. Object 2 for even-numbered animals and Object 1 for odd-numbered animals). On the final day (day 4), the last three steps were repeated however, instead of replacing one object with a novel object, one of the familiar objects was placed in a different location in the box. Similarly, the object moved in the box was changed for each animal to avoid bias. Statistical analysis. All samples were assigned randomly, and the experiments were performed in a singleblinded fashion. Specifically, the observer was unaware of the experimental conditions. Statistical significance tests were conducted with GraphPad Prism 8. One-way ANOVA was used to compare between multiple groups, followed by Dunnett's or Tukey's multiple comparisons tests for post hoc comparisons. Differences between the means of two conditions were tested using the two-sided Student's t-test with Welch's correction. Differences between data were considered significant if appropriate post-hoc statistical tests resulted in p ≤ 0.05. . The most suitable species were chosen for this study, and data were obtained in accordance with the best existing practices that ensure the animals' welfare 32 . The protocols used in this study are essentially identical to those employed in previous studies 33,34 and were adapted suitably. All in vivo experiments was carried out in accordance to ARRIVE guidelines.

Results
DEX exposure did not affect cell viability. In this study, we first sought to confirm whether various clinically used concentrations of DEX impacted cellular viability, and if these effects were dose-dependent. To avoid confounders of the whole animal or intact brain slices, we used a well-established in vitro cell culture model 33 . 1 to 2.5 μM has been recommended as a clinically relevant DEX concentration for in vitro experiments, and 5 μM deemed to be a higher dose in the clinical domain 39 . In the present study, we tested for the first time, all of the above concentrations to obtain a more comprehensive dose-response curve. Multiple areas (0.58 mm 2 ) of three replicates of each condition were imaged and quantified using one-way ANOVA testing, with identical thresholds at days 3 ( Fig. 1A-C) and 7 ( Fig. 1E-G) after DEX exposure. We found that DEX did not impact cellular viability at concentrations below 10 μM (Fig. 1). However, at 10 μM, DEX significantly reduced cellular viability, resulting in greater cell death (mean = 54.74%, SEM = 7.638 alive p = 0.0063, n = 3) compared to the control (mean = 82.3, SEM = 2.929, n = 3) at day 3 after exposure ( Fig. 1D). At day 7, DEX did not impact cellular viability any further when tested in all of the above concentrations (Fig. 1H). Collectively, these results suggest that DEX is non-cytotoxic-that is, it did not induce cell death in newly cultured, neonatal rat neurons at clinically relevant concentrations. Our dose response curve suggested that DEX at 1 μM was a safe and perhaps clinically relevant concentration, so all subsequent experiments from hereon were conducted with this concentration.
DEX administration promoted neuronal outgrowth. To assess whether DEX suppressed (like most other anesthetics) or impacted the total extent of neurite outgrowth, we tested its effect on developing rat cortical neurons in vitro. Specifically, we asked whether the growth of newly cultured neurons would be impacted by DEX. The cultured neurons were fixed and stained with polyclonal antibodies against the 160 kDa fragment of neurofilament (Fig. 2) and the extent of total neurite lengths was quantified in multiple areas of 0.15 mm 2 , using two-tailed t tests. We found that cells exposed to 1 μM DEX (Fig. 2C,D) had more neurite processes emanating from the cell body (mean = 5.274, SEM = 0.185, p = 0.0001, n = 13) compared to their control counterparts ( Fig. 2A,B) (mean = 3.864, SEM = 0.239, n = 12) (Fig. 2E). We also observed that the total neurite length per cell was greater for DEX treated cultures (mean = 298.0, SEM = 16.82, p = 0.0006, n = 13) as compared to their controls (mean = 211.9, SEM = 13.81, n = 12) (Fig. 2F). We did not however, observe significant differences in the average neurite length (control: mean = 55.74, SEM = 1.185, n = 12 and DEX: mean = 60.71, SEM = 2.492, p = 0.0895, n = 13) (Fig. 2G). Altogether, these results demonstrate that DEX did not affect neuronal ability to initiate growth, but rather cells exposed to this agent exhibited more branching per neuronal process as compared with their control counterparts.
DEX treatment did not impact synaptic formation. We next asked whether neurons exhibiting growth also established synaptic connections as defined by the juxtaposition of their presynaptic and postsynaptic proteins that are the hallmark of synaptic specialization. The cells cultured either in the absence or presence of DEX were fixed on day 7 and stained for two relevant synaptic proteins-synaptophysin (SYP) (Fig. 3A), a presynaptic vesicle protein, and postsynaptic density protein (PSD-95) (Fig. 3B). First, we assessed the relative levels of fluorescent intensity for both proteins in control and DEX-treated cells and analyzed the data using two-tailed t tests, Welch-corrected. We did not find significant changes in the expression levels for either SYP (control: Second, in order to determine synaptic specialization and formation of neuronal connections, we quantified the boutons where both pre-and post-synaptic proteins were expected to be in juxtaposition (Fig. 3C). Specifically, we measured the total number of synaptic puncta across the entire neurite length of a neuron, both in control   3F). Taken together, these data suggest that despite the increase in the length of neurite outgrowth/branching, there was no significant change in the total number of synaptic puncta per cell. These data demonstrate that, although DEX may have enhanced the length and the extent of growth, the total number of synaptic connections made between any two given cells did not change significantly.
DEX exposure resulted in a hyperfused mitochondrial network. Neuronal growth and synaptic assembly are high energy demanding processes that require intense mitochondrial efficacy. Previous studies have demonstrated that the mitochondria are a potential site for anesthetic-induced toxicity 5 , accordingly, we next asked whether DEX-mediated effects on cellular viability and growth may also have involved this organelle. In order to test the effects of DEX on mitochondrial morphology, we conducted live-cell imaging using MitoTracker Green to specifically monitor mitochondrial dynamics and subsequently analyzed this data using two-way ANOVA. Each individual neuron was assigned a distinct mitochondrial "identifier" based on their total numbers and the morphology that was specific to any given cell. This criteria for mitochondrial morphology characterization was based on previously established studies where they are classified either as fragmented when less than 0.75 μm in length, intermediate when 0.75-3 μm in length, and fused when greater than 3 μm in length 34 . We observed that the cells exposed to DEX had hyperfused mitochondrial networks at day 4 postexposure (Fig. 4C,D) (mean = 31.65%, SEM = 2.46 fused mitochondria, p < 0.0001, n = 10), compared to the control cells (Fig. 4A,B) (mean = 8.519%, SEM = 1.357 fused mitochondria, n = 8), and this was in complete contrast to what has previously been reported for other anesthetics 33,40,41 (Fig. 4E). In those instances where we did observe a fraction of cells showing fragmented morphology, these data were not statistically significant (control: mean = 20.03, SEM = 3.941, n = 8 and DEX: mean = 10.82, SEM = 1.520, p = 0.138, n = 10). To assess if this hyperfusion occurred in an acute manner, or was maintained over time, we analyzed mitochondrial morphology at days 7 and 10 post-exposure. We found that the hyperfused mitochondrial morphology was not maintained over time as by day 7 and day 10, neither was the fused nor the intermediate fraction of DEX-exposed cells were significantly different from the control (Fig. 4F). Taken together, these results demonstrate that DEX likely acts as a "switch" enabling the intermediate mitochondrial fraction to exhibit a more fused form in a time-dependent manner, which may likely enhance the energy production capacity of the cell mediating its positive effects. . When we compared side by side controls versus DEX exposed cells (Fig. 5C), we saw a significant difference in the MFI at day 7 between controls and DEX (control: D7 mean = 1613, SEM = 212.0, n = 4 and DEX: D7 mean = 1021, SEM = 117.2, p = 0.0087, n = 4). These data can be correlated with the results shown in the previous assay where we saw an increase in the hyperfused mitochondrial fraction in the DEX-exposed cells suggesting that the DEX treated cells continued to exhibit a healthy ROS production regime in a manner analogous to that observed during early stages of cultures.

DEX injected animals did not exhibit spatial or recognition learning and memory deficits.
To test whether DEX exposure compromised learning and memory in animals, rat pups were injected with a concentration previously described in the literature 22 and subsequently tested for learning and memory later in life. Specifically, we used P7 rat pups where 25 μg/kg DEX was administered subcutaneously either once (1×) or twice www.nature.com/scientificreports/ (2×) (second dose 24 h after the first dose). The animals were extensively monitored for their cardio-respiratory health, temperature, SpO 2 and righting reflex. These parameters did not differ significantly from the control animals injected with saline (data not shown) as previously shown in the literature 22 . At P60, the animals underwent MWM and NORT testing. For MWM, the animals were placed in a pool to reach a hidden platform invoking different spatial cues that were placed on the walls of the room for a period of 5 consecutive days (Fig. 6A). The average latency (time required to reach the platform once they were placed in the water) (Fig. 6B) and the distance traveled (Fig. 6C) before reaching the hidden platform were measured for each trial. One-way ANOVA statistical analysis was performed to compare the average latency and distance from day 1 to day 5. In terms of latency, we saw a statistically significant improvement in the time required to reach the platform comparing days 1 and 5 for both DEX-treated animals that were injected either once (1×) (Day 1: mean = 43.18, SEM = 3.325 n = 11 and Day 5: mean = 24.83, SEM = 2.722, p = 0.023 n = 11) or twice (2×) (Day 1: mean = 43.99, SEM = 3.557, n = 9 and Day 5: mean = 15.56, SEM = 2.695, p < 0.0001, n = 9) as well as controls (Day 1: mean = 46.6, SEM = 2.448, n = 16 and Day 5: mean = 20.14, SEM = 2.498, p = 0.006, n = 16) (Fig. 6B), suggesting that spatial learning had occurred in all groups. Regarding the distance swum by the animals, there was not a significant improvement between days 1 and 5 in the 1×-injected animals (Day 1: mean = 30.88, SEM = 6.356, n = 11 and Day 5: mean = 13.48, SEM = 3.073, p = 0.113, n = 11), which meant that this cohort of animals, in particular, swam longer distances in the pool to locate the platform as compared with saline and 2× DEX injected animals. Nevertheless, these animals still located the platform within a time window similar to that of the controls.  Taken together, these data show that DEX did not compromise learning and memory when administered to young pups, thus further underscoring its safety for clinical use.

Discussion
Most previous studies have shown DEX not only to be non-toxic but also neuroprotective against cytotoxicity induced by other anesthetics 17,18,20 . However, there exist controversies vis-à-vis its neuroprotective nature in situ 17,19 and also at the behavioral levels 18,19,42 . Moreover, the mechanisms also remain poorly defined, specifically in the context of its effects on neuronal growth and synaptic structure assembly at the level of the individual neuron-due mainly to the complex nature of the intact brain or intact brain slices used in other studies.
Although DEX can be used in combination with magnesium sulfate (MgSO 4 ) in the clinical setting 43,44 , we have only assessed DEX individually, as MgSO 4 has been reported to have neuroprotective properties 45 whereas we aimed to test DEX's properties on neuronal health on its own. This study provides the first direct and unequivocal evidence for the non-neurotoxic nature of DEX when tested for cellular viability, neurite outgrowth and synaptic network assembly. Moreover, we have demonstrated for the first time, that unlike other conventionally used anesthetic agents, which exert their neurotoxic effects by perturbing mitochondrial structure and function, DEX instead promotes mitochondrial health while maintaining a healthy ROS production. These results both endorse our previous finding that anesthetic-induced toxicity may involve the mitochondria while strongly suggesting that DEX-mediated positive effects on neuronal health may also involve this organelle.
Notwithstanding their essential requirements for most surgical procedures, the neurotoxic nature of various anesthetic agents, as deduced from numerous animal studies, cannot be overlooked. However, the investigation of the underlying neurotoxic mechanisms cannot be conducted in any human studies at the level of individual neurons-owing primarily to the experimental logistics and the ethical implications. While some studies do not support detrimental effects of anesthetics in humans 46,47 , a few others have suggested possible anestheticinduced toxicity in both developing 48,49 and aging brains 50,51 . These studies, together with research on non-human primates [52][53][54][55][56] , have raised significant red flags thus compelling healthcare regulatory agencies to recommend that warning labels be placed on anesthetics when given to pregnant mothers and young children 57 . Moreover, in recent years, several studies have urged that we take into consideration the potential neurotoxic effects of anesthetic agents on brain development in humans and define the underlying mechanisms using animal models 5,57,58 . Despite numerous attempts to understand the effects of anesthetics on neuronal homeostasis, the results remain either equivocal or controversial 5 .
In contrast to common conventional anesthetics, DEX is a newer agent being used both as a sedative and analgesic 59 agent that does not rely on NMDA or GABAA receptor function 60 . DEX acts on the alpha2 adrenoreceptor, acting as an agonist 59 , and some of its physiological functions include the induction of sedation, pain modification, inhibition of insulin release and the presynaptic inhibition of norepinephrine release 60 .
Although DEX is currently used in pediatric medicine 61,62 , there are also studies demonstrating that higher clinical concentrations of DEX indeed cause cell death in situ 25,63 . It would therefore appear that under certain experimental conditions DEX may induce neurotoxicity in brain slices and these subsequent effects may involve a variety of signalling pathways. However, in many other instances, DEX appears not only to be non-neurotoxic but also neuroprotective against the toxicity induced by other anesthetic agents such as sevoflurane, propofol or ketamine 22,64,65 . Previous studies have shown a number of anesthetic agents affect synaptic structures and function 33,66,67 ; however, similar conclusive evidence for DEX is missing. This study is thus the first to demonstrate that although DEX enhanced neurite outgrowth, it did not affect the total number of synaptic connections formed between the cortical neurons. When examined in light of previous studies where DEX-induced positive effects were shown to involve the PI3K/Akt/GSK3β pathway 18,20,42 invoking growth promoting factors (insulin for example), it therefore stands to reason that this anesthetic may activate signaling that could potentially be growth permissive. However, in our study, we found that notwithstanding this enhanced growth, neurons did not make additional synaptic contacts as compared with their control cohort.
Even though several studies have vouched for DEX safety, the evidence is not unequivocal. For instance, on the one hand, Liu et al. 25 tested different concentrations of DEX ranging from 0.001 to 0.2 µM, and found significant cell death in a dose-dependent manner. On the other hand, Sanders et al. 24 demonstrated that when cultured neurons that were previously treated with wortmannin and staurosporine (to promote neuroapoptosis) were exposed to 0.1-100 µM DEX, it rescued neurons from cell death in a dose-dependent manner 24 . In the present study however, we generated a dose response curve for DEX, ranging from 0.01 to 10 µM and identified a concentration of 1 µM to be the least cytotoxic. As this concentration is generally used in the clinic 39 , our data thus endorse it to be potentially safer in humans. Data presented in this study are also consistent with earlier work completed by Wu et al. which demonstrated that DEX protected neurons against apoptosis, axonal degradation, and synaptic degeneration after traumatic brain injury in a murine model 28 . Although Wu et al. 63 did not show direct effects of DEX on cellular viability, they did nevertheless demonstrate the extent to which it rescued neurons from axonal degradation and synaptic degeneration in the intact animal. The rationale for using www.nature.com/scientificreports/ DEX in a traumatic brain injury model used by Wu et al. 28 , where the induced damage is often inconsistent and unreliable remains unclear, however they did nevertheless find this anesthetic to be neuroprotective. The present study provides direct evidence that in a simple model system, which is devoid of whole animal complexity and inconsistent experimental paradigm (i.e. methods to induce traumatic injury), DEX prevented neurons from cell death and promoted neurite outgrowth (increased the neurite number and total length). These data thus confirm results obtained from in vivo studies and provide further unequivocal support at the resolution of individual neurons and synaptic structures. We suggest that these growth permissive or inductive aspects of DEX are likely attributable to its positive effects on mitochondrial structures, function and production of a healthy ROS regime. Notwithstanding the fact that DEX exerted growth permissive effects on neurites of cultured neurons, it did not however affect the total number of synaptic puncta between the two connecting neurons during the time window the cells were fixed (Day 7). It needs to be noted that most synapses that form between these neurons are thus consistent and remain unchanged even when cells are examined after 3 weeks. This argument is supported by several previous studies which showed that if in vitro neurons remained unperturbed due either the induction of electrical activity or pharmacological perturbations that the number of synapses formed between neurons remained unchanged for up to 100 days [68][69][70] . Whereas, at the cellular level, it would be almost impossible to demonstrate unequivocally that either the total number of persisting synapses, or the efficacy of an individual synapse would have remained consistent at any given time, an extensive study conducted at the proteomics level showed that the turnover of new synaptic proteins occurred primarily after the neuronal stimulation 71 . This rules out the possibility that DEX might tip the balance towards exuberant synaptic connectivity thus perturbing the innate balance of the synaptic "real-estate" that is shared between any two interconnected neurons. In contrast to these findings, previous studies have shown that both local and general anesthetics reduce the total number of synaptic puncta, as shown by a decrease in the relative expression of synaptophysin after isoflurane exposure 72 . It is therefore feasible that the neuroprotective effects of DEX observed in the Wu et al. 28 study may not have much to do with its role as an anesthetic, but rather it may have acted in a manner analogous to that of ketamine 34,73,74 . Specifically, their growth permissive or neuroprotective effects may have involved signaling mechanisms other then their modes of actions as an anesthetic agent 72 .
An important finding of this study is the link between DEX and its role in modifying mitochondrial networks. Specifically, we found that DEX exposure increased the fused mitochondrial fraction population in the neurons as compared with the intermediate fraction, without significantly modifying their fragmented form. Any prolonged imbalance in the fission/fusion equilibrium is widely documented to be detrimental to cellular health and viability, and has been shown to compromise key functions such as growth cone dynamics 75 , axonal regeneration, branching, and synapse formation 75-78 . Moreover, this study provided for the first time a comprehensive analysis regarding ROS production over time in cultured neurons. We found that DEX decreased ROS production over several days which could be the potential driver underlying its impact on neuronal health; this would however need to be tested experimentally in the future. Our data is nevertheless consistent with previous studies where DEX was shown to diminish ROS production in non-neuronal tissue such as kidney 79 and cardiomyocytes 80 , as well as following cerebral ischemia 81 .
As neurons used in our cultures were primarily from rat cortices, which we found to respond to anesthetics in a manner analogous to hippocampal neurons (data not shown), it could be argued that adrenergic receptors may have been absent and that this could account for the lack of evidence of DEX toxicity. To rule out this possibility, we injected rat pups for in vivo experiments, thus exposing the entire brain to DEX. These animals did not demonstrate any deficits in learning and memory, providing further support that DEX is non-neurotoxic when used in clinically relevant concentrations. Duan et al. 19 have previously demonstrated that when P7 pups (female only) were injected with DEX at a concentration similar to the one used in this study, their spatial memory tested in the Morris Water Maze 19 remained indistinguishable from the controls. In the present study, not only did we use one, but also two injections in a space of 24 h involving both male and female rats and invoked both Morris Water Maze and NORT learning and memory paradigms. The data presented here thus further endorses our in vitro results underscoring the non-neurotoxic nature of DEX and supports its utility in the clinical realm.