Introduction

The use of carbon dioxide gas (CO₂) for rodent euthanasia is accepted with conditions by the Canadian Council on Animal Care (CCAC) and the AVMA; however, CO₂ exposure is known to have adverse effects on both rodents and humans.^{1,4,11,40,41,42,44} CO₂ is aversive to rodents³⁴ at concentrations of 10.7 ± 1.1%,² and linearly activates nociceptors in the nasal mucosa of Wistar rats between approximately 37 and 87% CO₂ (≤ 25% CO₂ is subthreshold).³⁶ In humans, CO₂ exposure is reported as painful, with the lowest threshold for pain reported at a concentration of 32.5%.³ The CCAC and AVMA recognize that elimination of pain and distress during CO₂ exposure and euthanasia cannot always be achieved; however, CO₂ is viewed as a practical method that is cost effective and can be applied to many animals simultaneously.^1,11 For these reasons, CO₂ is frequently used for the euthanasia of rodents used in research in both European Union and Canada.⁶ However, an important issue is whether rodents lose consciousness before they experience pain during CO₂ euthanasia. Previously reported methods used for identifying loss of consciousness (LOC) vary considerably, including the use of outcomes such as recumbency, staggering and ataxia, and pedal withdrawal reflex.^{5,7,22,30,33,37,38} A broad range of CO₂ concentrations (21 to 39%) is associated with these endpoints.^5,7,22,33 This situation allows a potential overlap between CO₂ concentrations associated with LOC, nociceptor activation, and pain. Because pain cannot be experienced during unconsciousness,²³ the issue of whether LOC occurs at a CO₂ concentration below that associated with nociceptor activation is central to determining whether CO₂ overdose is a potentially painful killing method.

A popular proxy for identifying LOC in rodents is loss of righting reflex (LORR); with this method, a rodent is assumed to be unconscious if it fails to right itself after being placed in dorsal recumbency.^19,25,26,27 One report showed that LORR in rats and mice is strongly positively correlated with LOC in humans for a large number of different anesthetic agents.¹⁹ An investigation of the specific CO₂ concentration (threshold) that induces LOC in rats has not been performed using the LORR method.

The aim of the present study was to use LORR to investigate the concentration of CO₂ at which LOC occurs in rats. We hypothesized that the EC₉₅ CO₂ concentration for all strains would fall within the reported range for LOC during CO₂ exposure (21 to 39%)^5,7,22,33 and that rat strain and sex would have no significant effect on the EC₉₅. A secondary hypothesis was that the CO₂ concentration at which LORR occurs will be influenced by the LORR testing method used (1- or 4-Paw test).

Materials and Methods

Testing apparatus.

The testing apparatus was designed using software (SOLIDWORKS; Dassault Systèmes [SP1, Version 28], Waltham, MA) to create an operator-driven motorized rotating cylinder. Structural components were 3D printed with polylactic acid plastic filament using a 3D printer (Lulzbot Mini 2 3D printer; Fargo Additive Manufacturing Equipment 3D, Fargo, ND). A clear acrylic cylinder (outer diameter: 15.24 cm; inner diameter: 14.08 cm; length: 30.0 cm; and volume: 4.67 L) was suspended between 2-cylinder mounts (Figure 1) and rotated at a speed of 3 revolutions per minute. This rotation speed was selected based on similar previous reports in rats.^13,14,21,45

View Figure

• Download as Powerpoint
• Download Figure

A carrier gas O₂ was premixed with CO₂ before entering the chamber. To facilitate the control of gas concentrations, CO₂ and O₂ flow rates were individually adjusted by using 2 gas-specific, calibrated flowmeters (CONCOA Precision Gas Controls; O₂: 560 Series 150 mm Flowmeter; CO₂: 565 Series 65 mm Flowmeter; Virginia Beach, VA). The CO₂ concentration in the chamber was measured in real time using a CO₂ gas analyzer (CM-1000 Series 100% CO₂ Sampling Data Logger; GasLab, Ormond Beach, FL) in the gas outflow. The gas analyzer was calibrated each day before use. A red acrylic shield was placed between the observers and the apparatus to screen the rat from the observers.

Experimental design.

Before testing, rats were habituated to the experimenters, testing apparatus, and testing room over 5 d and considered habituated when readily interacting with experimenters and the apparatus environment. The experiment was conducted by 2 experimenters (DRM, BAM) who had different roles. The individual who controlled the apparatus, changed the CO₂ concentrations, and observed the rat’s behavior (DRM) was blind to the identity of the rat. The other person (BAM) performed randomization of the testing order and recorded all LORR scoring information and observations. The cylinder rotated continuously throughout the experiment, except during LORR testing. Cylinder rotations resumed between each LORR test.

Each rat was tested individually in one continuous experiment. Immediately after the rat was placed in the testing apparatus, cylinder rotation, and gas flow began at 100% O₂ (4 L/min) and 0% CO₂ for 8 min (baseline). The O₂ flow rate remained constant throughout the experiment. After the 8-min baseline period, the CO₂ concentration was increased to approximately 18% by adjusting the CO₂ flowmeter, guided by the CO₂ gas analyzer. A 6-min transition period was provided for CO₂ to reach the desired concentration in the chamber; this period was calculated based on preliminary testing to confirm gas equilibration. The transition period was followed by a 1-min hold period during which LORR testing was performed. After each hold period, the CO₂ concentration was increased by 1% or 2%. A 2% increase was used if the rat was ambulating, and a 1% increase was used if the rat was ataxic. A transition and hold period accompanied each change in CO₂ concentration. Experiments were video-recorded to confirm observations.

A LORR test was performed when cylinder rotation positioned the animal into dorsal recumbency, with all limbs in the air and no paws in contact with the cylinder. Once the rat was dorsal, cylinder rotation was stopped and a 15-s timer was started. The CO₂ concentration at the beginning and end of the 15-s period was recorded. If the rat had been incorrectly positioned at the beginning of the 15-s period, this was scored as false placement and rotations were resumed for repositioning. If a rat could not be positioned in dorsal recumbency at a given CO₂ concentration, the righting reflex was recorded as present at that concentration.

Two different LORR assessments were performed during each LORR test. In the 1-Paw assessment, LORR was achieved if the rat remained in dorsal recumbency with no paws in contact with the cylinder for 15 s. Conversely, if one or more paws came into contact with the cylinder during the 15-s period, then a failure to achieve LORR (righting reflex present) was recorded. The 4-Paw assessment was as described for the 1-Paw test, except that the righting reflex was considered present only if all 4 paws regained contact with the cylinder (LORR was considered to be achieved if 1 to 3 paws contacted the cylinder). These assessments were scored as a binary outcome (positive or negative). A positive LORR test constituted a rat remaining on its back for 15 s with 0 paws (1-Paw assessment) or < 4 paws (4-Paw assessment) moved to contact the cylinder. Conversely, a negative LORR test reflected a single paw in contact with the cylinder (1-Paw test) or all 4 paws in contact with the cylinder (4-Paw assessment) (Figure 2).

View Figure

• Download as Powerpoint
• Download Figure

LORR testing continued with incremental increases in CO₂ until 3 consecutive positive LORR tests occurred. After the final (third) positive LORR test, the CO₂ concentration was increased to 100% and the rat was euthanized without recovery. During video review, any LORR tests that had been performed incorrectly were excluded. Examples of this were if a rat was not properly positioned in dorsal recumbency (that is, rotated to one side or the other) or if a paw was in contact with the cylinder at the start of the 15s test period.

Results

The first rat tested (female Sprague–Dawley) was excluded from the data set because the testing methods were refined. In addition, 4 Sprague–Dawley females had been misidentified as males, resulting in unequal numbers of males (n = 10) and females (n = 17). No rats were excluded from the Wistar (n = 14 males, n = 14 females) or Long–Evans (n = 14 males, n = 14 females) groups. After the video review, 49 1-Paw LORR tests (15 Wistar; 9 Long–Evans; and 25 Sprague–Dawley) were excluded because LORR test criteria had not been met. In addition, 64 4-Paw tests (15 Wistar; 9 Long–Evans; and 40 Sprague–Dawley) were excluded for not meeting LORR test criteria. The total number of LORR tests per strain was as follows: Wistar, 295 1-Paw tests and 295 4-Paw; Long–Evans, 259 1-Paw tests and 259 4-Paw tests; and Sprague–Dawley, 282 1-Paw tests and 267 4-Paw tests. The data used to generate results are available in an online data repository: https://doi.org/10.7910/DVN/2CWEXH.

The 1-Paw EC₉₅ values (EC₉₅ [95% CI]) were significantly different between Sprague–Dawley (35.0% [34.0 to 36.5%]) strains as compared with Wistar (27.2% [26.3 to 28.4%], P = 0.0003) and Long–Evans (29.3% [28.0 to 31.5%], P = 0.011) strains but were not different between Wistar and Long–Evans strains (P = 0.516; Figure 3). A similar pattern was observed in the 4-Paw data, with the EC₉₅ for Sprague–Dawley (31.1% [30.5 to 32.0%]) significantly higher than both Wistar (26.2% [25.5 to 27.3%], P = 0.001) and Long–Evans (25.9% [25.1 to 27.4%], P = 0.0005) strains but no significant difference between the Wistar and Long-Evan strains (P = 0.971; Table 1; Figure 4). Male and female values showed no significant differences for any of the strains (Table 2). Comparisons of 1-Paw and 4-Paw data within strains showed a significant difference (EC₉₅; P = 0.010) for the Sprague–Dawley strain but not for the Wistar and Long–Evans strains (Table 3).

View Figure

• Download as Powerpoint
• Download Figure

View Figure

• Download as Powerpoint
• Download Figure

Discussion

The data from this study show that 1) the use of different methods to assess LORR (1-Paw compared with 4-Paw) can affect the results obtained; therefore, CO₂ concentrations associated with LORR can vary; 2) strains show significant differences in their susceptibility to CO₂ relative to LORR onset; and 3) males and females show no significant differences in their LORR responses to CO₂ regardless of strain.

Under the current guidelines for CO₂ euthanasia, death must be as quick and painless as possible^1,11 with consideration given to the animal’s experience.^9,28 The concepts of both nociception and pain are relevant with regard to LOC: neuronal nociception processing pathways are activated in both conscious and unconscious individuals, while pain is experienced only when conscious.²³ Therefore, an unconscious individual cannot perceive pain. This consideration forms the basis of the current study, which is to identify CO₂ concentrations associated with LORR as a proxy for LOC.

The effects of CO₂ on nociception and pain have been investigated previously in rodents and humans. Exposure to CO₂ elicits aversion behavior in adult, female Sprague–Dawley rats at concentrations as low as 10.7 ± 1.1%²; however, nociception and pain appear to occur at higher concentrations. In adult, male Wistar rats, most wide dynamic range neurons begin responding to CO₂ concentrations greater than or equal to 37%, with a linear increase in response until a plateau is achieved at approximately 87%.³⁶ Exposure of nasal mucosa to a CO₂ concentration of 25% was subthreshold for most neurons.³⁶ Although nociceptor activation cannot be assumed to equate to the presence of pain,²³ the overlap in CO₂ concentrations that are reported as painful in humans (≥ 32.5%)³ and associated with nociceptor activation in rats (≥ 37%)³⁶ and the 1-Paw EC₉₅ values observed in the current study suggest a low possibility of pain occurring in some individuals before LORR occurs. Ultimately, because evidence of CO₂ eliciting pain in rodents is indirect (in contrast to human self-reports), reliance on multiple lines of evidence is necessary.

LORR in rodents is positively correlated with unconsciousness in humans.¹⁹ A similar relationship has not been established for other endpoints used as measures of LOC in rodents. For example, recumbency occurs before LORR,^17,31 indicating that CO₂ concentrations associated with recumbence (21 to 39%) are unlikely to closely approximate LOC.^5,7,22,33 Limited research has been published on the concentration of CO₂ that induces LORR in rodents. These previous studies have measured the time to achieve LORR using CO₂ gradual fill rates, rather than the actual CO₂ concentration, as was done in the current study.^{17,30,31,32,39}

LORR scoring methods often use a binary behavioral response of the animal once it is placed in dorsal recumbence as either remaining in dorsal recumbence (indicating LOC) or returning to sternal recumbence or normal positioning (indicating consciousness).^19,25,26 Even though LORR is often assessed as a binary outcome, an array of specific behavioral observations have been reported.²⁹ For example, a variable number of paws in contact with the floor or anesthetic chamber surface has been used to distinguish between the absence or presence of LORR, as shown in other studies: “Each animal was placed on its back. The righting reflex was intact if the rat rotated onto all four paws within 10 s”¹⁸ (equivalent to the 4-Paw method of the current study), and “If the distal end of any limb made contact with the table surface, the righting reflex was considered intact.”⁴³ (equivalent to the 1-Paw method of the current study). As shown by the results of the current study, differences in LORR methodology affect the results.

A significant difference was found between 1-Paw and 4-Paw assessment methods in the Sprague–Dawley rats. The impact of methodology on results is important as it has the potential to affect data interpretation. The 1-Paw method is more stringent than the 4-Paw method, as reflected in the rightward shift in the dose-response curves for the 1-Paw data. With the 1-Paw assessment, a single paw in contact with the cylinder constituted the presence of the righting reflex, thus requiring a higher dose of anesthetic (in this case CO₂) to prevent paw contact. Furthermore, a recent study suggests that LORR is not a simple binary assessment but rather a spectrum of behaviors leading up to LOC.²⁰ Movement of a single paw (similar to the 1-Paw method used here) reflects a lower arousal state than movement of multiple paws, supporting the idea that the 1-Paw method is a closer approximation of LOC than the 4-Paw method. However, incomplete reporting of LORR test methodology and the variety of LORR test methods used²⁹ complicate direct comparisons of studies.

Our data reveal differences in strain-specific susceptibility of rats to CO₂. Sprague–Dawley rats exhibited LORR at higher CO₂ concentrations than Wistar and Long–Evans rats. Consequently, Sprague–Dawley rats appear to potentially require longer exposure to CO₂ before LOC occurs, with a greater likelihood of pain occurring before LOC: the Sprague–Dawley 1-Paw EC₉₅ was 35.0% (34.0 to 36.5%), which is above the lowest concentration reported as painful in humans (32.5%)³ and close to the concentration associated with nasal nociceptor activation by CO₂ in rats (37%).³⁶ This contrasts with the 1-Paw EC₉₅ CIs for Long–Evans (29.3% [28.0 to 31.5%]) and Wistar (27.2% [26.3 to 28.4%]) rats, which are below the lowest concentration reported as painful in humans and below the CO₂ concentration associated with nociception in rats.^3,36

We found no significant differences between male and female rats for the strains studied. Nonetheless, the consistently higher EC₅₀ and EC₉₅ values in females indicate that a small, but clinically unimportant, sex difference may exist. The reason for this sex difference is unknown.

This study was limited to the 3 rat strains studied. These strains were selected because they are among the most commonly used in biomedical research.²⁴ Although establishing EC₉₅ values for LORR in all rat strains is impractical, the possibility that some strains require higher concentrations of CO₂ to induce unconsciousness as compared with the strains studied here should be considered, and caution should be used when applying our results to other strains.

Another limitation is that we used a rotating motorized cylinder to place the rats in dorsal recumbency. Other studies have used other LORR placement methods such as positioning the animal by hand, tilting a chamber at a specified angle, or manually rotating a cylindrical cage by hand.²⁹ Comparing results between studies that use different methodologies is difficult, as highlighted by the 1-Paw and 4-Paw differences reported here. The use of oxygen as the carrier gas rather than medical air is also a potential factor. By preventing hypoxia, supplemental oxygen prolongs the time to death during CO₂ euthanasia.¹ However, this is quite different from the focus of the present study, which is to identify LORR. Nonetheless, because a comparison with medical air as a carrier gas was not performed, any impact of oxygen on the results cannot be ruled out. Finally, no statements can be made regarding aversion, anxiety, or distress, as these were not evaluated.