Client-owned dogs of both sexes from the breeds Greater Swiss Mountain Dogs and Border Collies, diagnosed with IE with or without antiseizure drug treatment and their not affected first degree relatives were enrolled into the study. Idiopathic epilepsy was diagnosed based on the International Veterinary Epilepsy Task Force TIER II confidence level (47). They were divided into newly diagnosed dogs without antiseizure drug treatment (IEU) and into dogs already under antiseizure drug treatment (IET). Idiopathic epileptic dogs were only included if they were scanned at least 2 days apart from the last recognized epileptic seizure to limit the possible influence of ictal changes on the spectroscopy (44). As a control group (control) not affected relatives (first degree relative to an epileptic dog but not clinically affected by IE) were included. All healthy relatives were followed up by telephone interviews at the time of writing the manuscript and are still free of seizures (minimum follow up period 2.5 years).

All dogs were fasted 8–12 h before anesthesia, with water being always accessible. In all dogs, medical history was collected, including information about administration of antiseizure drugs. All patients underwent a clinical and neurological examination performed by a board-certified neurologist. Furthermore, if no recent blood work was available, blood chemistry and hematology were obtained.

¹H-MRS was performed under general anesthesia. All dogs received 0.2 mg/kg butorphanol (Alvegesic 1% forte ad us. vet., Virbac, Switzerland) either intramuscularly (IM) or IV if a venous catheter (VasoVet, B. Braun Melsungen AG, Germany) had already been placed. The decision of placing a venous catheter into a cephalic vein before or after sedation with butorphanol depended on the behavior of the individual dog. In the Greater Swiss Mountain Dogs 1 mg/kg esomeprazole (Esomep 40 mg, Gruenenthal Pharma AG, Switzerland) was administered IV before induction. All dogs were preoxygenated for 5 min, before induction of general anesthesia with propofol (Propofol 1% MCT Fresenius, Fresenius Kabi AG, Switzerland). The dogs' tracheae were intubated with a cuffed endotracheal tube (Super Safety Clear, Teleflex Medical, Switzerland) of suitable size.

Anesthesia was maintained with sevoflurane (Sevorane, AbbVie AG, Switzerland) to effect, in oxygen and air. The dogs were mechanically ventilated with pressure—controlled ventilation (8–12 cmH₂O) through a circle system and respiratory rate was adapted to maintain end—tidal CO₂ between 35 and 38 mmHg. Animals were placed in dorsal recumbency during the ¹H-MRS, and the following parameters were measured continuously and recorded every 5 min throughout the procedure with a magnetic resonance imaging (MRI) compatible multiparameter monitor (Datex Ohmeda GE N-MRI2-01, Finland): peripheral oxygen saturation, heart rate, non-invasive blood pressure, respiratory rate, end-tidal CO₂, and end-tidal sevoflurane concentration.

All dogs received an infusion of Ringer's acetate (Ringer Acetate, Fresenius Kabi AG, Switzerland) at rates of 5 ml/kg/h IV. To maintain normotension, which was defined as a mean arterial blood pressure above 60 mmHg five ml/kg of Ringer's acetate were given IV over 10 min as bolus if needed. If up to four fluid boli of Ringer's acetate (5 ml/kg) did not increase the mean arterial blood pressure above 60 mmHg, a dobutamine (Dobutrex 250 mg/50 ml, Teva Pharma AG, Switzerland) constant rate infusion was started at a rate of 0.5 μg/kg/min to increase the mean arterial blood pressure above 60 mmHg.

MRI and MRS of the brain were performed with a 3 T MRI (Philips Ingenia, Philips AG, Switzerland) using a 15 channel receiver transmitter head coil (Stream Head-Spine coil solution, Philips AG, Zurich, Switzerland). Conventional morphological MRI was performed prior to the MRS acquisition. Single voxel MRS acquisition of the right or left thalamus with point resolved spectroscopy (PRESS) was subsequently performed as described in the study of Mauri et al. (31). Briefly summarized T2-W images in all 3 planes were used to place the single voxel in the thalamus, preferably in the right thalamus and carefully avoiding cerebrospinal fluid, as well as peripheral soft and bony tissues adjacent to the thalamus to prevent lipid contamination. Voxel size was 1.8 cm³ (10 × 12 × 15 mm). Details of the MRS protocol and analysis are given in the Supplementary Table S1.

After the first MRS acquisition, 1 mg/kg ketamine was administered IV and 2 min later the MRS scan was repeated. Prior to further analysis the quality of the spectra of the dogs were evaluated. Only spectra of dogs with a morphological normal brain, with correct voxel placement in the thalamus, and with adequate spectral quality (visual inspection for artifacts) were included in our study (Figure 1). The spectral quality in the final data set was assessed based on the signal to noise ratio and the full width at half maximum (FWHM) as given by the software used to analyze the spectra (LCModel, S Provencher, Oakville, ON, Canada) and the full width at half maximum (FWHM) from the unsuppressed water peak.

LCModel was used to estimate metabolite ratios to total creatine and water. Due to a variety of factors contributing to the signal (including not only voxel size, T1 and T2 relaxation times and repetition and echo times, but many more), these measured signals were calibrated against a commonly used reference metabolite (creatine) (48). The ratio of metabolite to total creatine was chosen because tissue segmentation required for correcting the ratio to water was not available in this study (31). Nevertheless, the ratios to water were used to check the results from the ratios to creatine. From all metabolites present in the basis set, only metabolites fitted with a median (over all measurements) relative Cramér Rao lower bound (CRLB) lower than 50 % were included in the following analysis (49).

A mixed ANOVA with the independent factors ketamine administration within subjects and group allocation (control, IEU, IET) between subjects was performed for all measured metabolites, heart rate and mean arterial blood pressure after checking assumptions for normality, outliers and homogeneity of variances and covariances. If significant effects were found, appropriate post-hoc tests were run with a Bonferroni correction for multiple comparisons. If the assumptions were not met, we performed a linear mixed effects model. A p-value < 0.05 was considered statistically significant.

Our study did not show any significant effects of ketamine or group allocation on GLX/creatine ratio in the thalamus. Glutamate as the major excitatory neurotransmitter is of special interest in epilepsy and an increased extracellular glutamate in the brain and / or a reduction in GABA concentrations are leading to excitotoxicity, epileptic seizures, and cell death (50).

Former MRS results for glutamate/creatine or GLX/creatine ratio after ketamine administration are limited and controversial, both for people and animals. Reported changes range from an increase to a decrease in glutamate or GLX levels. No changes were found in the GLX/creatine ratio in the thalamus of healthy awake volunteers 2 h after 0.8 mg/kg of ketamine IV administered over 50 min (51). Increased glutamate/creatine ratios were reported in the frontal cortex in healthy rats 30 min after a subanesthetic dose of 30 mg/kg of ketamine subcutaneously (52). Increased levels of glutamine were reported in the anterior cingulate cortex in humans after subanesthetic doses of 0.27 mg/kg ketamine IV followed by a constant rate infusion (CRI) for 2 hours (53). In contrast, decreasing GLX levels were measured in the medial prefrontal cortex of humans suffering from depression with increasing subanesthetic doses of ketamine (0.1–0.5 mg/kg IV) (35).

When looking at the canine literature, higher levels of GLX were reported in the striatum during general anesthesia with 15 mg/kg ketamine IV than during general anesthesia with 8 mg/kg pentobarbital IV (46). However, this study evaluated the absolute difference between these two types of anesthetic drugs in healthy dogs only. Thus, it is not clear if ketamine increased the GLX levels or if pentobarbital decreased them (46). Bektas et al. (manuscript in preparation) evaluated the acute effect of 2 mg/kg ketamine IV on GLX/creatine ratio in the in the hippocampus in healthy Beagle dogs under sevoflurane anesthesia and, in line with our results, did not find any significant changes in GLX/creatine ratios before and after ketamine administration. Comparisons to published studies are difficult due to differences in species and health status, dosage and administration modes of ketamine and variances in timing related to ketamine administration as well as differences in location of the MRS in the brain. There are some possible explanations why we could not detect significant changes in GLX/creatine ratio after ketamine administration.

Firstly, with the current MRS technics applied we were only able to measure GLX/creatine ratio, the sum of glutamine and glutamate. glutamine and glutamate are linked via the glutamine–glutamate cycle (54). Changes in the glutamine/glutamate ratios, with an increase in glutamine and a decrease in glutamate have been detected after NMDA—antagonist administration (55). Such changes in opposite directions after ketamine might stay undetected when only the sum of glutamate and glutamine can be measured. To differentiate between glutamate and glutamine a higher magnetic field strength or advanced MRS sequences would be needed.

Secondly, MRS is not able to differentiate between extra—and intracellular metabolites (56). Ketamine is expected to cause an increase of the extracellular glutamate levels (57). Furthermore, the total amount of glutamate is much higher intracellular than extracellular (58). Therefore, a possible extra—or intracellular shift of glutamate after ketamine administration cannot be detected with the current method. As the extracellular levels of glutamate determine the extent of excitation, separation of extra—and intracellular concentrations would be interesting, but currently no non—invasive technique is able to assess such shifts.

Thirdly, we investigated the effect of ketamine administration on the thalamic GLX/creatine ratios only. Since brain regions have different GLX/creatine ratios and different NMDA receptor distribution, it is possible that in other brain areas, changes in GLX/creatine ratios after ketamine administration could have been measured. More studies, with similar doses of ketamine and with voxel placement in different brain areas at higher magnetic field strengths could give us more insights into the effect of ketamine on glutamine and glutamate levels.

In a concurrent study investigating a similar group of dogs the MRS of the thalamus before ketamine showed decreased GLX/creatine ratios in the treated IE dogs compared to Controls and untreated IE dogs (31). The current study shows that, although changes in GLX/creatine ratio may be present in IET dogs, their reaction to a low dose of ketamine in general anesthesia seems comparable to their healthy and their untreated IE relatives.

The detected increase in glucose/creatine ratio in MRS could have two possible explanations. Firstly, ketamine could have increased the supply of glucose to the brain by increasing the blood glucose level, by decreasing the regional metabolic rate of glucose or by increasing the brain perfusion (59). Secondly, ketamine could have decreased brain glucose consumption (60). The decreased consumption is possibly caused by a direct NMDA blockade (54, 61) or indirectly by an increased anesthetic depth (62). Decreases in heart rate and blood pressure after ketamine in our study indicate deeper plane of anesthesia. Ketamine and deeper plane of anesthesia lead to vasodilation (63) with a possible concurrent increase in blood supply to the brain.

In people, a bolus of 0.5 mg/kg of ketamine IV showed reduction or increases of the regional metabolic rate of glucose in different brain regions (64). In contrast, a target-controlled ketamine infusion (0.3 μg/ml) led to an overall increase of regional metabolic rate of glucose in the whole brain, with the highest increase in the thalamus (65). However, another study revealed no change after a target-controlled s-ketamine infusion (0.75 μg/ml) of regional metabolic rate of glucose at all (66). In rats, high intraperitoneal dosages of ketamine (170 mg/kg) decreased regional metabolic rate of glucose in the thalamus in one study (67), whereas in another study 10–30 mg/kg of ketamine IV lead to different results according to the brain area, with no effect measured in the thalamus (68).

In human IE patients, glucose consumption is reduced inter-ictally and elevated ictally compared to baseline (69, 70). We did not detect different glucose/creatine response after ketamine administration between IE dogs and healthy control dogs. However, a trend for an interaction between group allocation and ketamine administration was found (Figure 2A) and the effect of ketamine on glucose/creatine ratio was most prominent in the IEU dogs. Interestingly, the dog with the reported seizure like phenomenon during recovery had the lowest increase in glucose/creatine. Whether this low increase of glucose/creatine ratio had been caused by an already ongoing silent epileptic seizure activity during the MRS or if it is a coincidental finding cannot be further elucidated.

Not only the regional metabolic rate of glucose, but also the blood glucose concentration can be changed by ketamine administration and can therefore be a reason for a change in brain glucose/creatine ratio. In rabbits, a low dose of 0.17 mg/kg of ketamine IV produced hyperglycemia and higher doses of 1–2 mg/kg IV produced hypoglycemia (71). In dogs, no elevation in blood glucose after a ketamine bolus of 1 or 2 mg/kg IV could be detected in one study (72). However, another study documented an elevation in blood glucose using 10 mg/kg of ketamine IV (73), so, a dose dependent elevation in blood glucose can be assumed. Unfortunately, blood glucose levels were not measured in the current study, which presents a limitation to discuss further what caused the elevated glucose/creatine ratios in the thalamus. Further studies are needed to explain the exact mechanism between ketamine and glucose/creatine ratio in the brain. Studying different dosages and different brain regions might give further insights into this finding.

Measuring GABA using ¹H-MRS is challenging due to the low concentration of GABA in the brain and the spectral overlap with other metabolites. Thus, GABA levels should be interpreted with caution, unless measured at higher magnetic field strength (≥7 T MRS) or special sequences (74). Even though we found a significant increase of GABA/creatine ratio in group IEU, the median relative CRLB were high, so the current results should be interpreted with care. In future studies the measurement of GABA in dogs could be optimized, although the voxel size is limited due to the small size of our patients' brain. Using spectral editing sequences such as MEGA-PRESS with longer scan times and correction for B₀ instability would enable more reliable GABA levels (75).

Ketamine administration blocks the postsynaptic NMDA receptor, therefore more glutamate can bind to pre-synaptic inhibitory glutamate receptors (76). When these receptors are activated by glutamate binding, a retrograde messenger is released, which can result in an increase in GABA (20, 77) and could further explain the increase in GABA/creatine ratio after ketamine administration. An increase in the GABA/creatine ratio could indicate a neuroprotective effect of ketamine in the epileptic dogs, by shifting the balance toward more inhibitory metabolites in the brain (12, 50).

Furthermore, lower GABA concentrations have been found in the urine in untreated IE dogs compared to treated IE dogs and healthy controls (32). Interestingly, ketamine reverses the GABA deficit in human major depression patients (35) which may also apply to untreated epileptic dogs. In epileptic patients, GABA turnover can be reduced, with not yet fully understood reasons (69). A reduced glycolysis and therefore a decrease in glutamate turnover in inter-ictal IE patients has been reported as a possible reason. As glutamate can be transformed into GABA via the glutamic acid decarboxylase, a decrease in glutamate turnover can lead to a decrease in GABA (69). It is possible that dogs treated with GABAergic anti-epileptic drugs, have GABA concentrations in the brain, that are more similar to concentrations in healthy individuals, which could be shown in studies in human medicine (41, 78, 79) and which is supported by the recent study from Mauri et al. (31). Furthermore, increases in GABA levels were reported in the medial prefrontal cortex after 0.5 mg/kg over 40 min of ketamine IV in humans suffering from obsessive—compulsive disorder (33). Possibly ketamine reverses the GABA deficit reported in untreated canine IE patients (51, 80).

Not only an increase in glucose/creatine ratio, but also a significant increase in ascorbate/creatine ratio was detected after ketamine administration. Ascorbate is an important antioxidant in the brain that can be reliably detected using MRS at 3 T field strength (81, 82) and resonates at 3.73 ppm which is close to the GLX compounds resonation (81). To preserve antioxidant effects, ascorbate underlies strict homeostasis in the brain (82). Increases in extracellular glutamate concentrations have been associated with increases in ascorbate concentrations in the brain under experimental conditions (83, 84). While the effect of ketamine on brain ascorbate levels has not been investigated, premedication with ascorbate before ketamine anesthesia has shown to enhance the effect of ketamine anesthesia in rabbits (85). In animal epileptic seizure models, administration of a high dose of ascorbate has shown anti–seizure and neuroprotective effects, but it is unknown whether ascorbate regulation is impacted by natural occurring epilepsy (86). No differences between groups have been found in the current study.

It is difficult to transfer the current findings into clinical implications as no studies comparing metabolite estimates and clinical effects in dogs are available. In this study we could prove that the effect of a drug on brain metabolites can be measured non-invasively using ¹H-MRS in clinical patients. This opens the door for future studies exploring the effect of antiseizure medication in epileptic dogs. Possible indication might be the investigation of drug resistance in epilepsy and to elucidate how ketamine interrupts refractory status epilepticus in dogs.

In the current study all animals were under general anesthesia and metabolite concentrations can be altered due to anesthetic choice of drugs (45, 87). It is therefore debatable if the findings after administration of ketamine under general anesthesia can be translated to the administration of ketamine in awake patients. To minimize the interaction of anesthetic drugs during the current study, we chose short acting drugs for premedication (butorphanol) and induction (propofol). For maintenance of anesthesia, we chose sevoflurane, as a recent study found that sevoflurane had only minor effects without clinical relevance on GLX concentrations (45).

We decided to use 1 mg/kg ketamine IV as this dosage is often used as co-induction agent and also for analgesic purposes (88). As we used a lower dose than the commonly used doses for status epilepticus treatment in dogs and because our dogs were scanned inter-ictally we cannot draw direct conclusions about the effects of higher dosages of ketamine in ictal dogs (10).

Antiseizure drug treatment can result in chronic changes in brain metabolite concentrations (89). The effects might differ between different drugs, different epileptic syndromes, and different brain areas (79, 90). The dogs included in our study were not on a standardized treatment and a variety of drugs in different dosages were used, precluding a definitive conclusion about the effect on antiseizure drug treatment on the MRS results.

As a further limitation electroencephalography (EEG) of the IE dogs was not available. Clinically no epileptic seizures were reported at least 2 days before the MRS scan for all IE dogs, but the absence of epileptic seizures was not confirmed by EEG and underlying epileptic seizure activity could have been missed. Additionally, we cannot prove the absence of any ictal activity during MRS scans, because of the lack of EEG during the MRS scan.

The volume of interest was chosen in the thalamus, due to the region's involvement in IE in humans (91) and dogs (92) and the presence of NMDA-receptors in the thalamus. As IE is not only involving the thalamus and ketamine is acting also in other parts of the brain (hippocampus, cortex), the effects of ketamine in other parts of the brain could be different and should be investigated in future studies by examining different brain regions.

The sample size was relatively small with groups not balanced regarding breeds, age and sex. However, the sex distribution with more males than female in IEU and IET dogs reflects the natural distribution of the disease in dogs, with a higher risk of epilepsy in male dogs (93, 94). Further studies assessing differences among breeds are needed to exclude a possible bias.

Some dogs were treated with dobutamine to maintain non-invasive mean arterial blood pressure above 60 mmHg. However, no change of rate before and after the ketamine administration was performed, so a possible influence of dobutamine would not interfere with our study results.

Increase of GABAergic transmission can be also due to acute or chronic stress (95). Dogs are commonly stressed in veterinary environment (96), but we did not investigate stress levels of the individual dogs to account for stress as a possible cofounder.