The collagenase model of intracerebral hemorrhage in awake, freely moving animals: The effects of isoflurane

Intracerebral hemorrhage (ICH) is a devastating stroke often modelled in rats. Isoflurane anesthetic, commonly used in preclinical research, affects general physiology (e.g., blood pressure) and electrophysiology (e.g., burst suppression) in many ways. These physiological changes may detract from the clinical relevance of the model. Here, we revised the standard collagenase model to produce an ICH in rats without anesthetic. Guide cannulas were implanted stereotaxically under anesthetic. After 3 days of recovery, collagenase was infused through an internal cannula into the striatum of animals randomly assigned to the non-anesthetized or isoflurane group. We assessed whether isoflurane affected hematoma volume, core temperature, movement activity, pain, blood pressure, and seizure activity. With a small ICH, there was a hematoma volume increased from 8.6 (± 3.3, 95% confidence interval) µL in anesthetized rats to 13.2 (± 3.1) µL in non-anesthetized rats (P = 0.008), but with a larger ICH, hematoma volumes were similar. Isoflurane decreased temperature by 1.3°C (± 0.16°C, P < 0.001) for 2 h and caused a 35.1 (± 1.7) mmHg group difference in blood pressure (P < 0.007) for 12 m. Blood glucose increased twofold after isoflurane procedures (P < 0.001). Pain, as assessed with the rat grimace scale, did not differ between groups. Seizure incidence rate (62.5%) in non-anesthetized ICH rats was similar to historic amounts (61.3%). In conclusion, isoflurane appears to have some significant and injury size-dependent effects on the collagenase model. Thus, when anesthetic effects are a known concern, the use of the standardized cannula infusion approach is scientifically and ethically acceptable.


Introduction
Intracerebral hemorrhage (ICH) has a mortality rate of approximately 40%, and accounts for 10-20% of all strokes (Feigin et al., 2009). Despite improvements in patient care and risk factor management, incidence and mortality rates have not changed in the past three decades (van Asch et al., 2010). In order to enhance the chance of translational success and replicability, preclinical animal models must be as accurate to clinical presentation as possible. Currently, a major discrepancy of many preclinical ICH studies is the Therefore, the effects of anesthetics must be carefully considered in preclinical research, including when testing neuroprotective agents (Seto et al., 2014).
Despite the promising neuroprotective potential of isoflurane, it may also impact stroke outcome negatively, especially during or following ICH. As mentioned, isoflurane is a potent vasodilator that increases cerebral blood flow (Constantinides and Murphy, 2016;Sicard et al., 2003). During a hemorrhage, an increase in cerebral blood flow could increase the size of the hematoma; however, this may be compensated for, as isoflurane also reduces blood pressure (BP) and heart rate (Yang et al., 2014). In addition to these hemodynamic factors, temperature is an important component in hemostasis (Wolberg et al., 2004). Volatile anesthetics can lower temperature by several degrees in patients (Khurram et al., 2014) and animals (Colbourne et al., 1993) during and transiently after surgical procedures. This is potentially problematic, as mild reductions in temperature can inhibit clotting enzymes and can cause platelet dysfunction (Wolberg et al., 2004). Temperature reductions also impact other mechanisms of injury, such as inflammation and apoptosis (Moore et al., 2011). Thus, establishing whether anesthetic influences these physiological factors in the collagenase model of ICH is vital, as the predictive accuracy of our preclinical model could be affected.
As with spontaneous ICH in patients, rats experience seizures in the days following a collagenase-induced ICH (Klahr et al., 2016(Klahr et al., , 2014. Electroencephalograms (EEG) show burst suppression under isoflurane anesthesia, a pattern in which cortical activity alternates between silence and high amplitude slow or sharp waves (Amzica, 2009). Isoflurane also disrupts cortical network connectivity (Hentschke et al., 2017), a feature that has also been associated with seizures (Vega-Zelaya et al., 2015). However, isoflurane also depresses glutamate transmission, which is believed to play a role in post-ICH injury (Castillo et al., 2002). How these alterations influence seizure activity following the collagenase ICH model is unclear; therefore, establishing whether isoflurane is interacting with the injury to produce seizures, or whether this is an intrinsic property of the ICH model is important. Post-stroke seizure activity may be notably different in frequency and intensity in animals that are not given anesthetics.
Thus, in order to assess isoflurane's impact on the collagenase ICH model, we utilized a common drug delivery approach to infuse collagenase into awake animals. Rats were randomized to either the isoflurane (ISO) or the no isoflurane (NO-ISO) group. In the first experiment, we assessed the impact of isoflurane on blood glucose as well as hematoma volume, temperature, activity, and pain acutely following a small ICH (Fig. 1). In the second experiment, based on these results, we again assessed hematoma volume, blood glucose, and pain after a large ICH to determine if our findings were dependent on ICH size. In the third experiment, we compared the historic characteristics, including rate, duration, and severity of post-stroke seizures commonly observed following an anesthetized ICH procedure with that seen here in conscious animals to determine the effect of isoflurane on seizure activity. In the fourth experiment, we compared the BP and hematoma volume observed during and following the anesthetized and conscious ICH models to investigate one mechanism by which isoflurane may be affecting hematoma volume.

Exclusions and mortality
Five animals were completely excluded from experiment 2, and 1 additional animal was excluded from RGS analysis. One animal in the ISO group died spontaneously after ICH, and 1 animal in the NO-ISO group was not given an ICH due to a blocked cannula. In the ISO group, 1 animal was excluded from RGS analysis because he was sleeping for the duration of the video. Three animals were excluded based on insufficient hematoma volume, 2 from the ISO group and 1 from the NO-ISO group. In experiment 3, one animal spontaneously died~42 h after ICH, but was still included in assessment. Two animals' BP data in experiment 4 were excluded from the ISO group due to blocked catheter. No animals were excluded or prematurely euthanized for animal welfare concerns.
2.2. Experiment 1: The effects of isoflurane after a small ICH 2.2.1. Blood glucose Blood glucose was measured before and after the telemetry probe and cannula implantation surgeries, in which both groups were exposed to isoflurane. There was a significant interaction between group and time on blood glucose after the telemetry probe and cannula implantation surgeries (Fig. 2c, P = 0.046, interaction effect). There were no group differences at either time of measurement ( Fig. 2c, P = 0.489, before surgery; P = 0.300, after surgery). Blood glucose was increased by the end of the surgery in both groups (Fig. 2c, ISO group P < 0.001, Cohen's d = 1.91; NO-ISO group P < 0.001, Cohen's d = 2.83).

Hematoma volume
Hematoma volume was significantly larger in the NO-ISO group ( Fig. 2a, P = 0.042, Cohen's d = 0.60, moderate effect). Data was not distributed normally in either group (both P < 0.05, Shapiro-Wilk's test). When assessed with a Mann-Whitney U test, the NO-ISO group still had a significantly larger hematoma volume than the ISO group ( Fig. 2, P = 0.008).

Temperature and activity
Average core temperature during the 24 h baseline did not differ between the groups (P = 0.088, data not shown). There was a significant interaction between time and group on temperature post-ICH ( Fig. 3a, P < 0.001, interaction effect). Isoflurane significantly reduced temperature for the first 2 h after surgery (Fig. 3a, P < 0.001 immediately post-ICH, Cohen's d = 3.62; P = 0.037 at 1 h post-ICH, Cohen's d = 1.10). Similarly, there was a significant interaction between time and group on activity (Fig. 3b, P = 0.020, interaction effect). The NO-ISO rats were less active at 10 h post-ICH only (Fig. 3b, P = 0.037, Cohen's d = 1.17). As this difference only occurred at one time comparison, it is likely due to chance.

Rodent grimace scale
There was no group difference in observed pain after ICH at either time ( Fig. 4a, P = 0.577, 6 h; P = 0.529, 23 h). Pain scores did not differ between 6 and 23 h post-ICH ( Fig. 4a, P = 0.333). There were no differences in pain between groups on any of the four subscales (all P > 0.454). The median score of 0.28 shows the majority of rats are below the proposed analgesic intervention threshold of 0.67 (Oliver et al., 2014). In the ISO group, 1 rat was above this threshold, and in the NO ISO group there were 3 rats above this threshold (P = 0.609).

Weight loss
There was no difference in weight loss between the groups (P = 0.602, interaction effect; P = 0.685, group main effect) and there was no significant weight loss after the ICH (P = 0.659, time main effect, data not shown).
2.3. Experiment 2: The effects of isoflurane after a large ICH

Blood glucose
Blood glucose levels were significantly increased after cannula implantation surgery ( Fig. 2d, P < 0.001, time main effect, partial η 2 = 0.79). There was no difference in blood glucose between the ISO and NO-ISO groups ( Fig. 2d, P = 0.824, group main effect; P = 0.807, interaction effect), who were both given isoflurane for that procedure, which was three days prior to the ICH induction.

Hematoma volume
Isoflurane did not significantly impact hematoma volume in this experiment ( Fig. 2b, P = 0.169). Animals in the NO-ISO group had significantly less variability in hematoma sizes as compared to the ISO group ( Fig. 2b, P = 0.009). Our conclusion did not change when analysis was redone with Welch's correction (P = 0.183).

Rodent grimace scale
There was no effect of isoflurane on rat grimace scale scores after ICH ( Fig. 4b, P = 0.358, 6 h; P = 0.936, 23 h). There were no group differences in pain on any of the four subscales (all P > 0.151). With the larger ICH, the median score of 0.80 is slightly higher than the analgesic intervention score of 0.67, indicating that approximately half of the animals were experiencing pain (Oliver et al., 2014). The median score seen here is similar to what is seen in other injuries, such as spinal cord injury (Schneider et al., 2016), and lower than a previous ICH study (Saine et al., 2016). The grimace scale scores did not improve or worsen by 23 h post-ICH ( Fig. 4b, P = 0.333).

Weight loss
There was significant weight loss after the ICH (P < 0.001, time main effect), with animals losing an average of 5.7% of their body weight. However, there was no difference in weight loss between the groups (P = 0.991, interaction effect; P = 0.537, group main effect, data not shown).
2.4. Experiment 3: Seizure activity after non-anesthetized ICH 2.4.1. Baseline EEG in the collagenase group When we compared the RMS of slow wave sleep of the day prior to stroke with days 1 and 2 post-ICH in those rats with epileptiform activity, we found a significant effect for day (P = 0.013) but no interaction with hemisphere (P = 0.266, Fig. 5a). This shows that the fluctuations in non-epileptic EEG traces after ICH were higher on day 1 compared to day 2 in both hemispheres.

Seizures after collagenase
Five out of eight rats (62.5%) had seizures and interictal spikes (Figs. 5-7) within the first two days after the stroke, with the earliest occurring after~7.5 h and the latest occurring after~45 h. Seizures ranged in duration from~5 to 45 s (Table 1, Fig. 6). All rats with seizures had extended periods of abnormal interictal activity, as seen previously (Klahr et al., 2016(Klahr et al., , 2014 Table 2. Most seizures were bilateral, but there were instances in which seizures occurred only ipsilaterally or contralaterally (Fig. 6). Surprisingly, in animals that spent the least amount of time in seizure activity, slow wave sleep RMS was reduced compared to baseline slow wave sleep RMS, indicating that the fluctuations had decreased in amplitude. Increases in power were seen at almost all frequencies for rat 3 (Table 1), which had the highest RMS and the longest seizures. However, for rat 1, which had a low RMS, it had an overall decrease in power, with those frequencies mainly affected being those below 11 Hz. For coherence calculations, we concentrated on increases in coherence at the frequencies in which we noted a change in power. Coherence values range from 0 to 1; values of 0 indicating that signalspecific frequencies between the two channels are completely unrelated, whereas values of 1 indicate that they are completely related. We found that for animal 1 there was an overall increase in coherence. Even though rat 3 had power increases along most frequencies of interest, the seizures showed mostly decreases in coherence in frequencies ranging from 0.12 to 1 Hz and 6-13 Hz, and 20-25 Hz and increases in frequencies between 40 and 50 Hz.

Lesion volume
The mean lesion volume in this experiment was 47.1 mm 3 ( ± 31.8 mm 3 , 95% CI). We did not observe a relationship between lesion volume and number of seizures ( Fig. 5b, R 2 = 0.170, P = 0.310) or the total duration of seizures (R 2 = 0.103, P = 0.438).
2.5. Experiment 4: Effect of isoflurane on blood pressure and hematoma volume 2.5.1. Blood pressure During the ICH procedure, there was a significant interaction between group and time on BP ( Fig. 8a, P = 0.003), but these changes were transient. The animals in the NO-ISO group had significantly elevated BP as compared to the ISO group from minutes 2-12 of the ICH procedure ( Fig. 8a, all P < 0.007, partial η 2 = 0.43). Note that animals in the ISO group were kept under isoflurane for 25 m, but animals in the NO-ISO group were typically done their infusion procedure within 10 m (infusion plus handling time).
In the 24 h following the ICH procedure, there was no effect of group on hourly averaged BP ( Fig. 8b, P = 0.492, group main effect; P = 0.152, interaction effect). There was a significant effect of time on BP ( Fig. 8b, P < 0.001), likely a circadian rhythm effect.

Hematoma volume
There was no group difference in hematoma volume (P = 0.955, Blood glucose was assessed immediately before and after surgery, while hematoma volume was assessed 24 h after small ICH induction. Awake ICH rats had significantly larger (a) hematoma volumes (µL; P = 0.0082), and both groups had higher (c) post-surgical blood glucose levels (mmol/L; P < 0.001) compared to pre-surgery, with significant group by time interactions (P = 0.046). One rat in this experiment had a negative hematoma volume, likely due to variability in vasculature blood between hemispheres, as hematoma volume is calculated as ipsilateralcontralateral blood volume. Experiment 2. After a larger ICH, isoflurane had no effect on (b) hematoma volume (P = 0.183), though awake rats had less variability in hematoma size observed (P = 0.009). Both groups had (d) elevated blood glucose levels following cannula surgery (P < 0.001) and were not significantly different (P = 0.824). Sample sizes were n = 24 per group in experiment 1 and n = 18 per group in experiment 2. ISO = 12.61 ± 11.7, NO-ISO = 13.05 ± 15.69, data not shown). The size of the hematoma was not predicted by peak BP (R 2 = 0.097, P = 0.382), average BP (R 2 = 0.244, P = 0.147), average BP during the ICH procedure (R 2 = 0.103, P = 0.366), or average BP during the first 6 h post-ICH (R 2 = 0.252, P = 0.139).

Discussion
Models of ICH typically require anesthetics that often markedly affect physiology (Tsurugizawa et al., 2016). Here, we used a cannula infusion system to investigate a model of collagenase ICH in non-anesthetized animals in order to assess the effects of isoflurane on outcome after ICH. Non-anesthetized animals had similar pain levels after ICH, indicating that this is an ethically acceptable way to induce an ICH and avoid anesthetic confounds. Isoflurane slightly decreased BP, decreased temperature, and increased blood glucose; however, these effects often resolved within minutes to hours. The effects of isoflurane lead to an increase in bleeding when the ICH was small, but did not significantly increase bleeding when the ICH was large. It is possible that other anesthetics or a longer duration of anesthetic may have greater effects, although Esposito and colleagues found that higher doses of isoflurane did not impact collagenase induced ICH (Esposito et al., 2013). This cannula infusion model is beneficial in cases where anesthetic effects are being tested and when anesthetic interactions with other therapies are a concern.
We used the RGS to determine if the lack of isoflurane during the ICH procedure resulted in additional pain. We found that the RGS scores both 6 and 23 h after ICH did not differ between groups, indicating that conducting the procedure under isoflurane anesthesia does not reduce post-ICH pain and is an ethical method of ICH induction. Further, animals experienced approximately the same weight loss in the day after ICH, providing further evidence that the non-anesthetized procedure was not causing significant additional stress. This also suggests that most of the pain experienced after collagenase-induced ICH is due to the injury itself and not the surgical procedure. After a large ICH, the majority of rats in both groups were above the analgesic Isoflurane rats had (a) significantly lower core temperature (°C) for 2 h (P < 0.001 at 1 h, p = 0.0373 at 2 h) and (b) lower activity at 11 h (AU = arbitrary units of animal movement detected and recorded by a telemetry receiver placed under the animal (Colbourne et al., 1998), p = 0.0372) post-ICH induction. There was a group by time interaction for both (a) temperature (P < 0.01) and (b) activity (P = 0.020). Sample size was 12 per group. Rat grimace scale (RGS) scores were not significantly different between isoflurane and awake rats after (a) a small ICH (P = 0.369) or (b) a moderate to large ICH (P = 0.402). Pain was also not significantly different between 6 and 23 h post-ICH induction for either experiment 1 (P = 0.421) or experiment 2 (P = 0.435). Sample size was 24 per group for experiment 1 and 18 per group for experiment 2.
intervention threshold (Oliver et al., 2014). In this case, we did not treat the pain in these rats, as selectively treating animals with analgesics that influence injury can confound studies (Ferland et al., 2007;Saine et al., 2016). Others may choose to treat post-stroke pain, and further research is needed to determine the impact of those analgesics on bleeding and brain injury.
As hypothesized, isoflurane affected all of the physiological variables that were tested, and would have affected many others that were not tested. Despite that, we found only little differences between models, likely because these effects were short lasting or minor in nature. Temperature was decreased for the first 2 h post-surgery in the isoflurane group, and was increased above baseline in the conscious group. However, in this case, such a transient and modest temperature difference only led to a significant difference in hematoma volume after a small ICH. In the NO-ISO group, BP increased for the duration of the ICH procedure, and quickly returned to baseline once the dummy cannula was replaced. Anesthetized animals had their BP gradually drop throughout the duration of the surgical procedure, which returned to baseline within the hour. We expected that fluctuations in BP would result in large differences in hematoma volume. The lack of effect may be due to the collagenase enzyme causing prolonged bleeding over several hours (MacLellan et al., 2008), whereas the observed BP changes only lasted for up to 25 m. Additionally, BP increases may not have equated to increases in cerebral perfusion pressure. Isoflurane increases intracranial pressure (Scheller et al., 1988), and after large ICH, intracranial pressure increases even further (Hiploylee and Colbourne, 2014;Silasi et al., 2009). When intracranial pressure increases and arterial BP decreases by a similar amount, cerebral perfusion pressure stays the same (Su et al., 2008), which may be one possible explanation for why we only see an increase in hematoma volume after a small ICH. The observed BP changes may have effects that we did not measure. For example, stress responses, such as increased corticosterone, may also affect the course of injury and recovery (DeVries et al., 2001). Clinically, ICH patients are commonly hypertensive, and BP is often elevated during the course of injury (Shah et al., 2007), although this is typically for longer and to a much greater extent than what we see in rats. Therefore, the more accurate model of BP during an ICH would likely be increased BP.
Isoflurane increased blood glucose by nearly twofold by the end of   Table). (b) Seizure of rat 3 with high power and high RMS. (c) Example of contralateral seizure occurring in rat 1.
the surgical procedure. We could not determine how long blood glucose was elevated for, as blood sampling required anesthetizing the animals or additional handling stress. However, previous research shows that in mice, isoflurane-induced glucose elevations resolve within an hour (Durand et al., 2009). In patients, higher blood glucose at admission was associated with mortality (Fogelholm et al., 2005). In a rat model of focal ischemia, the harmful effects of hyperglycemia are well known (Kagansky et al., 2001). It is possible that the non-anesthetized rats experience a stress induced hyperglycemic response during the ICH procedure, as indicated by increased blood-pressure (Márquez et al., 2004). As many preclinical ICH studies use a similar anesthetic protocol as we use here, we believe that a substantial portion of ICH research involves acute hyperglycemia at the time of the bleed. In animals without anesthetic, we observed a post-ICH seizure rate of 5/8 (62.5%), which is similar to past work that found 61.3% of rats had seizures (Klahr et al., 2016(Klahr et al., , 2014 but higher than the rate seen clinically and in the autologous whole blood model of ICH (De Herdt et al., 2011;Klahr et al., 2014). With the NO-ISO, we saw seizures limited to the contralateral hemisphere in two animals, which has not been seen in our previous work. One animal had decreased coherence during a bilateral seizure, suggesting two independent seizures were happening at the same time. Following ischemic stroke, seizure foci may be present contralaterally as well as ipsilaterally, and can propagate via the thalamus (Kogure and Schwartzman, 1980). Isoflurane can affect thalamic information transfer (Detsch et al., 1999), thalamic GABA A receptors (Jia et al., 2007), and blunt the overall thalamic response to stimuli (Vahle-Hinz et al., 2007), which may be the reason we do not see contralateral seizures after animals have been exposed to isoflurane. Additionally, in most cases, the power change during seizures compared to baseline was quite small, suggesting that these seizures were not as severe as in previous work (Klahr et al., 2014). As isoflurane suppresses neural activity and prevents seizures in some cases (Bar-Klein et al., 2016;Ito et al., 2012), it is possible that after isoflurane exposure, animals may experience rebound effects, such as exacerbated seizure activity (Veronesi et al., 2008). Therefore, isoflurane may be altering the characteristics of seizures rather than the frequency that seizures occur. However, because this experiment used historical controls rather than an experimental control group, results should be interpreted with caution, and follow up experiments should be done.
This study had some additional limitations. We did not assess the long-term effects of isoflurane on injury and behaviour (Jiang et al., 2017;Seto et al., 2014;Statler et al., 2006). Future studies should examine chronic effects of isoflurane after ICH. Further, we did not assess the effects of isoflurane in the autologous whole blood model, as it would be technically difficult to do. Although the physiological impacts are likely similar, our conclusions are limited to the collagenase model of ICH. Here, we did not visually assess seizure activity, as we did not video record any of the seizure events. Therefore, we cannot provide any conclusive information regarding the behavioural manifestations of this epileptiform activity.
Our findings suggest brief use of isoflurane anesthesia is an appropriate model. However, we recommend the use of non-anesthetized ICH in specific cases. For example, isoflurane may interact with many drugs (Glass et al., 1997;Wood, 1991), including analgesics (e.g., opiates) and potential neuroprotectants, especially those given before or at the time of surgery. Validating findings in a model without anesthesia would be beneficial to ensure effects are not dependent on or interacting with isoflurane. Further, avoiding anesthetic would be ideal in situations where researchers are specifically examining the effects of Left-Example of seizure in rat 1 displayed lower power (top) but increased coherence (bottom) than baseline SWS EEG, meaning the magnitude of brain activity was decreased while coordination between hemispheres was increased. Right-Example of seizure in rat 3 displayed higher power and lower coherence, signifying that the magnitude of epileptiform activity was increased, and activity in each hemisphere was less coordinated than baseline levels. The dashed lines in the power spectrum represent the 95% CI and the solid lines represent the mean values for the spectrum. For the coherence, any increase in coherence above the CI limit of the difference between seizure and baseline SWS was significant.
anesthetics. Previously, most studies looking at the effects of anesthetics on stroke simply compared anesthetics to each other, or compared different doses of the same anesthetic, without having a no anesthetic control group (Bhardwaj et al., 2001;Zausinger et al., 2002). By using a non-anesthetized ICH, researchers can compare findings to a no anesthetic control group, increasing the validity of their findings. Although we only assessed isoflurane in this study, a variety of other anesthetics are commonly used in ICH research (e.g., chloral hydrate and sodium pentobarbital, MacLellan et al., 2012) and should be similarly assessed for their effects on ICH outcome.
In conclusion, we demonstrate that isoflurane can have small hematoma-size-dependent impacts on the ICH model. The physiological impacts of isoflurane are often transient, but should still be considered a potential confound. The non-anesthetized ICH is an important tool for researchers when anesthetic effects are a concern, especially in neuroprotectant studies and those with a short survival time.

Subjects and exclusion criteria
All procedures were done according to the Canadian Council on Animal Care Guidelines and were approved (protocol AUP960) by the Biosciences Animal Care and Use Committee at the University of Alberta. We obtained 104 male Sprague Dawley rats (275-600 g, approximately 2-4 months old) from Charles River (Saint Constant, Quebec). Animals were kept in a temperature and humidity-controlled room with lights on from 7:00 am to 7:00 pm, and all procedures were done during the light phase. Animals were single-housed during experiments with food and water provided ad libitum. All animals were handled for a total of 30 m over days prior to the collagenase infusion to decrease stress during handling. Rats were handled gently including repeatedly touching the dummy cannula to give rats exposure to having the device manipulated and to increase comfort level with experimenter handling.
We established a priori exclusion criteria of a hematoma volume < 5 μL for experiment 2, thus excluding any animals that did not receive a moderate-sized ICH.

Experimental design
Animals were randomly assigned to groups using random.org and data was analyzed in a blinded manner for all experiments. Group sizes were calculated a priori using a power analysis for a desired 80% power to detect a 33% difference in the primary endpoint, hematoma volume. In this study, we conducted 4 experiments to examine the effect of isoflurane on collagenase-induced ICH (Fig. 1). Animals were either randomized to receive an ICH under ISO or NO-ISO to examine the effect of isoflurane on ICH. In experiment 1, we assessed whether isoflurane affected temperature and activity for the first 24 h after a small collagenase-induced ICH. Blood glucose was measured at the beginning and end of anesthetic procedures, and pain was measured at 6 and 23 h post-ICH. Hematoma volume was assessed at 24 h post-ICH. Rats were randomized to ISO (n = 12) or NO-ISO (n = 12) conditions. As the variability was greater than expected, this experiment was repeated with an additional 12 animals per group to increase statistical power and the results were pooled for a total sample of n = 24 per group. Temperature and activity were only assessed in the first 24 animals (n = 12 per group). In experiment 2, we assessed the impact of isoflurane on blood glucose before and after surgical procedures, pain at 6 and 23 h post-ICH, and hematoma volume at 24 h after a large collagenase ICH. We used a larger insult in this experiment to test whether the effects of isoflurane were dependent on bleed size. Here, we had 18 animals per group. In experiment 3, we determined whether isoflurane influenced seizure activity after stroke. Here, we assigned 8 animals to the NO-ISO group. After 48 h of EEG activity measurements post-ICH, Table 1 Characteristics of seizures after NO-ISO ICH. EEG activity occurring in the first two days after collagenase injection were visualized and analyzed. For each rat, the number of seizures, laterality, total duration, and time of onset were documented. We also reported other factors to determine the variability of the traces as depicted by the RMS ratio (RMS seizure/RMS non-epileptic activity) and changes in voltage according to their frequency, as depicted by power changes and frequencies affected. Also, for those frequencies in which the power was affected, coherence was assessed as an indicator of how coupled the activity was in between both channels. Here, coherence was computed as an increase from baseline coherence. Traces shorter than 25 s were not analyzed for power and coherence. we collected brain tissue for lesion volume assessment. We determined post-ICH seizure incidence rate, which is~61.3% after anesthetized collagenase-induced ICH (Klahr et al., 2016(Klahr et al., , 2014. In experiment 4, we determined the effects of isoflurane and the no-anesthetic procedure on BP and hematoma volume after ICH. Animals were randomly assigned to ISO (n = 6) or NO-ISO (n = 6).

Telemetry probe implantation
Rats were anesthetized with isoflurane (4% induction, 2-2.25% maintenance, 60% N 2 O, and remainder O 2 ). In animals used for core temperature measurements, a sterile calibrated probe was inserted into the peritoneal cavity (Model TA10TA-F40, Data Sciences International, St. Paul, MN, accurate to ± 0.2°C). In animals used for EEG analysis, an EEG telemetry probe (F40EET, Data Sciences International, sampled at 500 Hz and low-pass filtered at 100 Hz) was inserted into the peritoneal cavity as previously described (Klahr et al., 2014). Leads were channeled under the skin and attached to screws (B000FMWBA0, Small Parts) and secured with dental cement. The screws were placed ipsilateral (AP −4, ML −4) and contralateral (AP-4, ML 4) to Bregma to avoid interfering with the cannula. In animals used for BP measurements, a calibrated PA-C10 probe's catheter was inserted into the left femoral artery, and the probe (Data Sciences International, accurate to ± 3 mmHg) was implanted subcutaneously, as previously described (Hiploylee and Colbourne, 2014).
A guide cannula was implanted (see Cannula Implantation and Table 2 Characteristics of seizures from our previous studies (Klahr et al., 2014(Klahr et al., , 2016. In both studies, the lesion volume averaged 33 mm 3 . Note that survival times were longer in these studies, ranging from 11 to 66 d. However, first seizure incidence was always within the first 24 h.  Awake animals had significantly higher BP from 2 to 12 m of the ICH surgery when compared to the isoflurane group (P < 0.007). Isoflurane animals were kept under anesthesia for 25 m, while the infusion procedure was typically completed in awake animals within 10 m. No group effect was found on (b) BP 24 h after ICH induction (mmHg; averaged hourly, P = 0.4924, interaction effect P = 0.1524). There was a significant effect of time on BP (P < 0.001). Sample size was 4 in the isoflurane group and 6 in the awake group.
Collagenase Infusion) following implantation procedures. Meloxicam (0.2 mg/kg SC) and bupivacaine hydrochloride (0.5 mg/kg SC) were administered for analgesia, with the exception of EEG implantation where only bupivacaine hydrochloride was administered. This was to avoid excessive analgesia and suture removal by animals, which otherwise would be more problematic with this procedure. Baseline measurements were taken for 24 h prior to ICH induction. Data measurements were taken every 30 s prior to and after the ICH. Post-ICH core temperature and activity readings were corrected hourly to baseline values in order to account for temperature changes due to circadian rhythm. In our past experience, we do not see spontaneous seizure activity in this rat strain (Klahr et al., 2016(Klahr et al., , 2014. Nonetheless, baseline EEG data was assessed to ensure there was no pre-ICH seizure activity. For BP measurements, baseline data was taken for 3 h on the day prior to ICH and averaged. All data were corrected for probe offset readings taken prior to implantation.

Cannula implantation and collagenase infusion
Animals were anesthetized with isoflurane and temperature was maintained at 37°C using a rectal temperature probe and heating pad placed under the animal. Meloxicam and bupivacaine hydrochloride were administered at the start of surgical procedures as an analgesic. A hole was drilled into the skull at 0.5 mm anterior, 3.5 mm lateral (left side) to Bregma (Paxinos and Watson, 2014). The dura mater was punctured to minimize possible pain during the non-anesthetized procedure. A guide cannula (C316G/SPC guide with 1 mm below pedestal, Invivo1, Roanoke, VA) was placed onto the hole and secured in place using 3 anchoring screws and dental cement. Dummy cannula was placed on the guide cannula to prevent pathogen entry and to maintain patency during recovery.
Rats recovered for 3 days following the cannula implant procedure. Then, an internal cannula (C316I/SPC, 5.5 mm extension from guide, invivo1, Roanoke, VA) was inserted 6.5 mm into the striatum (the depth used in previous research). Bacterial collagenase (Type IV-S, Sigma, 0.6 U/μL in sterile saline) was infused through PE tubing into the internal cannula and striatum. Either 1.0 μL (experiment 1) or 3.0 μL (experiments 2-4) of collagenase solution was infused over 2.5 m. The internal cannula was kept in place for 5 m to prevent backflow before being slowly removed, and the dummy cannula reinserted. Animals in the NO-ISO group were awake and lightly restrained by the experimenter during the cannula insertion. The experimenter used her hands to briefly hold the rat's head still for~10 s. Animals were allowed to freely move within the procedure space during infusion after the infusion cannula was inserted. Animals tolerated this procedure well, with no visible signs of discomfort. This infusion technique is commonly used in neuroscience research, including research on anxiety (Shah et al., 2004;Williams et al., 1987). Animals in the anesthetized group were kept under isoflurane anesthesia for exactly 25 m including an~2 m induction time (the duration of a typical collagenase procedure in prior research) with temperature maintained as described above. Note that the cannula implant procedure 3 days prior also took approximately 25 m of anesthetic.

Blood glucose measurements
Blood glucose was measured with a glucometer immediately after anesthetic induction and again prior to the end of surgical procedures (Contour Next One, Ascensia Diabetes Care, Mississauga, ON). A small drop of capillary blood was obtained from the tail by needle prick. Rats were free feeding prior to surgical procedures.

Grimace scale
Pain was assessed using the rat grimace scale (RGS) to test whether the conscious ICH procedure caused additional pain post-ICH (Sotocinal et al., 2011). Animals were video recorded for 10 m at both 6 h and 23 h post-ICH. We did not assess pain during the procedure, as collagenaseinduced bleeding occurs over hours (MacLellan et al., 2008). Ten images were selected from each video at approximately 1 m intervals (to allow full view of face at each time). Each image was scored on orbital tightening, nose/cheek flattening, ear changes, and whisker changes by a scorer blinded to group identity. Each subscale was scored from 0 to 2, and the combined score was added for a total score ranging from 0 to 8, with zero meaning no pain. Group differences between both subscales and total scores were assessed. For ease of comparison with previous research, total scores were averaged by subscale for a score ranging from 0 to 2. Animal's scores were the average of the scores on the 10 images. A second rater scored a subset of animals to check for inter-rater reliability, which was high (Spearman's rho of 0.839, P = 0.004). Four naïve animals served as a negative control, and these animals received an average score of 0.162.

Hemoglobin assay
The amount of hemoglobin in each hemisphere was determined using a spectroscopic assay based upon a standard curve (MacLellan et al., 2008). Hematoma volume was calculated as ipsilateral blood volume minus contralateral blood volume. This accounts for the blood in the hemisphere that is not attributed to the hematoma (blood in the vasculature).

Electroencephalogram analysis
Baseline and post-infusion EEG traces were visualized with Dataquest A.R.T software (v. 2.3, Data Sciences International). Two to five-minute long epochs of slow wave sleep (SWS) activity prior, and day 1 and 2 after collagenase injection as well as putative epileptiform traces were exported and analyzed using custom code written in MATLAB (R2012a, Mathworks, Natick, MA). We computed the root mean square (RMS), a measure of the fluctuation in the EEG signal, for all traces of interest. Baseline SWS from the day prior to collagenase injection (control) and putative epileptiform traces were exported and analyzed using custom code written in MATLAB for detection of seizures. The code identified epileptiform peaks that were above 4 standard deviations from the mean of the control traces, and considered 10 peak clusters occurring within 1 s apart as seizures (Klahr et al., 2016(Klahr et al., , 2014. Seizures detected by the MATLAB code were also visually verified for further accuracy, and any artefacts were excluded. Power spectral density using Welch's averaged modified periodogram method (6 s window, 2 s overlap) as well as dual channel coherence for ipsilateral and contralateral recordings (3 s window; 1 s overlap) were computed for those epileptiform traces longer than 25 s. Power spectra and coherence were compared to non-epileptic control traces similar in duration to determine which frequencies had significant changes in power during seizure activity (i.e., outside 95% confidence interval, CI). A randomized coherence distribution based on a series of sequential time-shifted (and also time-reversed) coherence computations from these actual traces was computed to calculate the coherence significance level. During epileptiform activity, we determine cross-hemispheric coupling changes by subtracting the coherence values of normal activity from those of epileptic traces and considered that any increases equal to or larger than the confidence limit for that trace to be significant.

Histological analysis
At 48 h post-ICH, animals from the seizure experiment were injected with 100 mg/kg IP of sodium pentobarbital and perfused with 0.9% saline followed by 10% neutral-buffered formalin. Prior to cryostat sectioning, brains were placed in a 20% sucrose solution for cryoprotection. Coronal sections (20-μm thick) were taken every 200 μm and analyzed every 400 μm. Sections were stained with cresyl-violet as done previously (MacLellan et al., 2006). Due to edema confounds and distortion caused by the hematoma itself at 48 h post-ICH, we were not able to calculate lesion volume by comparing hemisphere volumes (Williamson and Colbourne, 2017). Instead, the lesion was calculated as: (average area of damage × interval between sections × number of sections).

Statistical analysis
Data were analyzed using GraphPad Prism (v 6.0, GraphPad Software Inc., La Jolla, CA). Data are presented as mean ± 95% CI. Ttests were used for 2 group comparisons. Welch's correction was applied when group variances were not equal, as determined with an F test to compare variances. Blood glucose levels and RMS were analyzed using a 2-way repeated measures ANOVA with Sidak's multiple comparisons test. Rat grimace scale scores were assessed using a Mann-Whitney U test, and group contingencies were assessed using a Fisher's exact test. Relationship between variables was assessed with Pearson's correlation coefficient. All P values below 0.05 were considered statistically significant.
Author's contributions CMW, ACJK, and FC designed the experiment. CMW, ACJK, TK, DRM, and ACK collected the data. CMW, DRM, ACK, and CTD analyzed the data. CMW, ACJK, TK, and ACK wrote the manuscript and all authors edited the manuscript.

Availability of data
All electronic data from this study are available on Mendeley data.

Funding
This work was supported by the Heart and Stroke Foundation of Canada (Grant #G-16-00014065). Frederick Colbourne is supported by a Canada Research Chair award. The funding sources had no involvement in study design, data collection, data analysis, or the decision to publish.