Responses of the in vitro turtle brain to visual and auditory stimuli during severe hypoxia

Despite the abundance of oxygen in the atmosphere, its availability for aerobic respiratory function can vary considerably for animals, particularly for aquatic species, where oxygen solubility in water is low, competition for it can be high and its exchange with the atmosphere limited. When oxygen levels fall (hypoxia) to undetectable levels (anoxia), survival relies entirely on the ability of that animal to utilize anaerobic pathways for energy production, which, in vertebrates, yields just a fraction of the ATP per mole of glucose compared with aerobic respiration. Consequently, surviving anoxia is highly dependent on an animal's capacity to decrease its metabolic rate.

Among the champions of vertebrate anoxia tolerance are several species of North American pond turtle (Emydidae), which can survive anoxia for as many as 5 months at 3°C (Odegard et al., 2018) and for 36 h at room temperature (Johlin and Moreland, 1933). This performance would be impossible without a greater than 70% decrease in metabolic rate, which is achieved through the arresting of protein synthesis, ion channel function and synaptic transmission (Jackson, 2002; Buck and Pamenter, 2018). In turn, a decreased circulatory requirement further enables suppression of cardiovascular function (Hicks and Farrell, 2000a,b), producing additional energy savings. Extreme anoxia tolerance has been exploited by some overwintering pond turtle species (i.e. painted turtles, Chrysemys picta) that spend their winters in ice-locked freshwater ponds and bogs that become severely hypoxic and anoxic for many months across their native geographic range.

Although the importance of hypoxia- or anoxia-induced metabolic suppression is without question, lowering cellular respiration produces obvious functional trade-offs for tissue or organ systems. Presumably, during the anoxic episode, specific tissues and organs that are either non-essential or energetically expensive should be the first to exhibit functional and metabolic suppression, while essential and energetically inexpensive systems will be the last affected. Nervous system function is considered energetically expensive, and predictably exhibits functional suppression, particularly in the cerebrocortex (Buck and Pamenter, 2018; Buck et al., 2012). Less well explored are other brain regions such as the brainstem, where homeostatic pathways might continue to respond to external environmental cues during the winter. In the laboratory, one of us (D.E.W.) has observed that anoxic turtles can be aroused and will explore their anoxic environment, which naturally leads to basic questions about the biology of these hypoxia-tolerant animals: how are responses in sensory pathways altered by oxygen deprivation, and do those changes affect the way these animals interact with their environment?

The experiments described herein focus on the well-studied special senses processed in the brainstem that mediate responses to light and sound. External cues, such as photon absorption by retinal photoreceptors and mechanical vibrations of the tympanic membrane, are necessary for normal sensory function, but are their responses maintained at a similar level during anoxia? A previous study in turtles (Stensløkken et al., 2008) showed decreases in all of the electroretinogram (ERG) waveforms during anoxia. But the transduction mechanisms of the auditory system are extremely different. Phototransduction is energetically expensive, requiring continuous consumption of high-energy phosphates to maintain the signaling cascade that regulates cyclic-nucleotide gated sodium channel function in the photoreceptors. Auditory function, however, involves auditory mechanoreceptors for which energy consumption occurs in more discrete periods that track the physical displacement of the hair cells. To test the hypothesis that these basic differences in signal transduction mechanism translate into functional differences during severe hypoxia, we used reduced in vitro preparations of red-eared slider turtles (Trachemys scripta elegans), bathed in either oxygenated or severely hypoxic artificial cerebrospinal fluid (aCSF), and recorded field potentials when visual and auditory stimuli were presented. We report that neural responses do, in fact, differ for these two sensory systems during severe hypoxia.

Red-eared pond sliders, Trachemys scripta elegans (Wied-Neuwied 1838) (both sexes, 175–250 g), a relatively anoxia-tolerant North American pond turtle, were purchased from Niles Biological, Sacramento, CA, USA. They were housed in aquaria filled with municipal water (18–22°C) with access to basking platforms illuminated by halogen flood lamps (Sylvania 90 W aluminized reflector) and fed Aquatic Turtle MonsterDiet (Zeigler Bros., Gardners, PA, USA) three times weekly, ad libitum. All turtle procedures were in strict adherence to the Guide for the Care and Use of Laboratory animals of the National Institutes of Health and approved by the Institutional Animal Care and Use Committee of Saint Louis University (protocol 2979).

Following induction of surgical anesthesia (ice for approximately 30 min and then an injection of propofol, 10–20 mg kg⁻¹) into the neck's sagittal subcarapacial sinus), the head was decapitated and then bathed in ice-cold, oxygenated aCSF (also referred to as perfusate or medium) for subsequent dissection. The composition of the aCSF used throughout the study was always the following (in mmol l⁻¹): 96.5 NaCl, 2.6 KCl, 2.0 MgCl₂, 4.0 CaCl₂, 31.5 NaHCO₃, 274 mOsmol. When bubbled with 5% CO₂, this produced a pH of 7.6 at room temperature (Kogo and Ariel, 1997). Except when severe hypoxia was imposed, the aCSF was always bubbled with 95% O₂/5% CO₂. During severe hypoxia, the perfused medium was simply the control medium bubbled with CO₂ (usually 5%, but in some instances, 3%) with the gaseous balance always being N₂. There were no differences in responses between the two types of media, so the results from these experiments were combined for analysis and presentation.

The neural tissues were placed in a recording chamber perfused with oxygenated aCSF and fixed to a vibration-isolation table (Vibraplane, Kinetic Systems, Boston, MA, USA). Optodes were placed in the perfusing solution in the chamber (Witrox from Loligo Systems, Viborg, Denmark) to measure the P_O2 continuously, while a recording micropipette was visually guided to its neural target. Three preparations were employed to investigate the two forms of sensory stimulation, as described below.

The preparation used to study auditory function involved recording evoked potentials from the cochlear nuclei in an in vitro preparation. Those nuclei are visible in the medial brainstem, after the overlying cerebellum was deflected laterally, revealing them in the cerebellar peduncle adjacent to the vestibular nuclei (Fan et al., 1997). Field potentials were recorded during tympanic membrane (TM) vibration that was then relayed into the temporal bone, transduced by hair cells and transmitted as spike activity to the brainstem via the eighth cranial (vestibulocochlear) nerve.

Drawings and photographs of the TM and the reduced preparation are shown in Fig. 1, along with examples of field potential responses recorded in the brainstem. A TM was identified by two concentric wrinkles in the skin on this lateral photograph of a living turtle on the lateral wall of its head, caudal to the eye and mouth, just ventral to its signature patch of red skin (Fig. 1C). Fig. 1A shows the head structure drawn from above. A probe mounted on a piezo actuator pressed on the turtle's left TM, which is connected via an ossicle (thin rectangle) that transverses in the air-filled middle ear (gray oval) to the transducing organ (otic capsule, a small white oval) in the temporal bone.

To prepare this in vitro preparation, the cranium was hemisected along the mid-sagittal plane through the base on the cranium to generate a half-head preparation because those responses are best found ipsilateral to the stimulated tympanic membrane. The rostral bones in front of the temporal bones (the jaw, palate, orbit and nose) and the rostral telencephalic brain and its adjacent eyes and olfactory nerves were then removed. The remaining preparation was then hemisected, leaving the brainstem attached to the lateral head, with its labyrinthine sensory organs intact and functional. These tissues were secured, dorsal side up, by a brass strap mounted across the recording chamber to ensure their physical stability during independent movements of the nearby flexible tympanic membrane (Fig. 1D, see also Fig. 2A,B). Parts of the midbrain, cerebellum and spinal cord remained connected to facilitate physical stability of the neural tissue by pinning these brainstem structures to the Sylgard-coated floor of the recording chamber (Sylgard-184 Silicone Elastomer, Dow Corning, Midland, MI, USA). Extra care was taken not to grasp the temporal bones by the bilateral bony depressions that are the two tympanic membranes, because they are each connected to their underlying thin columella bones that relay vibrations of each tympanic membrane to each otic capsule. Damage to the delicate columellae and the flexible hydraulic seals of their end-feet onto the oval windows of each otic capsule might damage the relay of TM vibrations, not to mention causing harm to the sensitive stereocilia of the auditory hair cells therein.

The tympanum was displaced with a smooth glass probe mounted on a piezo-electric motor (PZL-007 linear actuator, Burleigh, Fishers, NY, USA). The probe was fashioned from a 2-mm glass capillary tube to form a narrow tip that was sealed by fire-polishing. It was then placed gently on the central region of the TM so as not to damage the underlying mechanical auditory structures (Fig. 1A,E). The tip position of the glass probe was adjusted by a micromanipulator to touch the skin surface covering the TM itself, while visually observed through a high-magnification dissecting microscope (Olympus SZX12 Stereo Microscope, Center Valley, PA, USA). The piezo-electric linear actuator moved the probe up to 7-µm step movements that were roughly orthogonal to the surface of the tympanic membrane as driven by a PZM 150 V driver amplifier (Burleigh). The stepper is rated to have a 5% non-linearity, 16% hysteresis and a 4.5 KHz frequency response. A Master-8 Pulse Stimulator (AMPI, Jerusalem, Israel) controlled the driver of the piezo actuator.

To calibrate the excursion of the stimulating probe that pushed on the TM, a stiff tungsten rod was fixed to the piezo-electric actuator in place of the almost transparent glass probe and aligned on a reticle on a microscope stage beneath a high-magnification microscope. The linear position of this opaque rod was photographed at the extremes of piezo steps (Fig. 1B). The tip of the tungsten rod, analogous to the glass probe, transversed ∼7 μm, as expected from the PZL-007 piezo-electric linear actuator. Preliminary experiments indicated that mechanical steps that pushed onto a central position of the TM would evoke a brainstem field potential at fractions of the maximal distance of 7 µm (also see Christensen-Dalsgaard et al., 2012). This range of movement is consistent with the dimensions of turtle hair cells of slider turtles, which exhibit stereocilia as tall as 10 μm that are estimated to be deflected roughly 100 to 200 nm (Nam et al., 2019; Spoon and Grant, 2013).

To verify that the field potentials were responding to the TM movements, voltages were recorded in the lateral brainstem at or near the region of the cochlear nuclei. The final placement of the micropipette tip was then positioned by a micromanipulator during piezo steps of the TM every 3 s to identify a location that evoked large, time-locked deflections in the voltage trace (e.g. red arrowhead in Fig. 2B). In general, the response onset latency was a few milliseconds and the response amplitude of the averaged field potential increased in a graded manner as the TM displacement amplitude increased or the stimulus duration increased. The size and shape of these evoked averaged field potentials were measured as a function of the movement parameters evoked by the piezo-electric pusher onto the tympanic membrane (see Fig. 1F–H). These brainstem responses to the glass probe steps demonstrated the physiologically normal aspects of the brainstem auditory responses, i.e. that slider turtles are known to respond to sounds within both aquatic and terrestrial environments, with higher auditory sensitivity relative to turtles that live only aquatic or terrestrial lives (Zeyl and Johnston, 2015). Consistent with hair cell mechanoreception in other species, TM movements in these experiments responded to high-frequency stimulation with very short latencies (Fig. S1), have waveforms that differ based on TM movement direction (Fig. 1G), and have larger response amplitudes for increasing movement amplitude (Fig. 1H). At stimulus amplitudes of 7 µm, responses do not diminish with repetition, so appear not to damage the stereocilia.

The procedures used for recording the in vitro visual responses have been described in several publications, for which neural recordings were made in visual brainstem structures including pretectum, the nuclei of the accessory optic system, the optic tectum (OT), the nucleus isthmi and the cerebellum (Ariel and Fan, 1993; Ariel and Johny, 2007; Fan et al., 1995; Kogo et al., 1998; Rosenberg and Ariel, 1990; Saha et al., 2010). The present study examined only the brainstem responses from the optic tectum. In brief, once the jaw and attached muscles were removed, the eyelids, conjunctivae and bone that formed the orbits were dissected away from each eye. The cranium was opened and the brain and its attached eyes were removed from all the bones. The entire telencephalon was cut from the thalamus and discarded. Eyecups were formed by slicing each eye along its anteroposterior midline, followed by the removal of the lens and excess vitreous humor. The entire preparation of both eyecups attached to the brainstem was submerged in control medium in a recording chamber, with both the retinas facing upward, filled with oxygenated medium and exposed to the diffuse bright-white light from a light-emitting diode (LED; driven by a 5-V pulse from the output of Digidata 1320A controlled by ClampEx 9 software, Axon Instruments, Foster City, CA, USA).

These experiments used LED pulse durations as a measure of stimulus light intensity. Stimulus duration was controlled by voltage steps with great accuracy given the temporal nature of the nearly instantaneous onset and offset of LED light pulses. Phototransduction, in contrast, proceeds very slowly, especially for poikilothermic animals such as aquatic turtles spending their winters in near-freezing environments. Therefore, turtle photoreceptors integrate individual photon absorptions as transduction events within a long time window of hundreds of milliseconds for cold conditions, especially for rod photoreceptors. At warmer temperatures, integration times may be as short as 100 ms for cone photoreceptors (Baylor and Hodgkin, 1973; Granda et al., 1986).

For recording of the field potential responses in the brainstem, the OT of the dorsal midbrain was placed upward in the recording chamber, and its overlying meninges were carefully peeled away to permit a recording micropipette tip to smoothly penetrate the tectal surface to a depth of approximately 0.5 mm. The field potentials recorded there represent the aggregate, low-pass filtered spike activity near the electrode tip in the tectum (Fig. S4).

In some cases, the surgery first created the eye and brainstem preparation, while preserving the caudal head tissue for use in a separate preparation that included the tympanic membranes and the temporal bones as described above. The various preparations were secured into plexiglass chambers containing a continuous flow of oxygenated perfusate. The brainstem was affixed by several pins to the Sylgard floor of the recording chambers (see details in Rosenberg and Ariel, 1990). During each recording session, a wide range of stimulus intensities was presented to determine the range of intensities (0.05–10 ms) that elicited electrical field responses that were consistently above response threshold and below response saturation. Light intensities within that range (1–5 ms) were selected to monitor the visual responses during the normoxic and severe hypoxic aCSF conditions.

Prior to the onset of severe hypoxia, control responses were recorded to establish a stable baseline for the sensory responses under oxygenated conditions. Then, a stopcock located close to the recording chamber switched the perfusate from the oxygenated medium (95% O₂/5% CO₂) to severely hypoxic medium (95% N₂/% 5% CO₂). To minimize contamination of the perfusate with atmospheric oxygen, the tubing carrying the medium from the reservoir to the recording chamber had low oxygen permeability (Versilon™-c-210 tubing; Saint-Gobain, Akron, OH, USA) or PTFE tubing ensheathed in an outer tube filled with flowing N₂ gas. In this OT preparation, as with preparation 1 for auditory recordings, the recording chamber was covered with a thin cover-glass to block air diffusion into the chamber's perfusate. After completing brainstem recordings, the overlying manipulators and cover-glass were removed to permit unobstructed overhead photography.

To isolate the tectal responses from those of the retina itself, effects of severe hypoxia on the ERGs from single eyes were determined by visually evoked ERGs produced in intact eyecups. Unlike intact eyes, where the corneal/vitreal side of the retina is electrically isolated from the scleral side of the retina, ERGs recorded after the eye was hemisected to form an eyecup (Fig. S2B) created a potential conductive path that could shunt the radial retinal current between the two sides of the retina if submerged. This shunting path was minimized within the special eyecup recording chamber and the cap for that chamber as follows.

In vitro ERG recordings require that the eyecup be placed in a gaseous environment so that the field potential of the retina's inner surface is electrically isolated from its scleral surface. To this end, a special eye chamber was fabricated to support the eyecup vitreal side up in a humidified gaseous environment. The chamber was covered by a cap with several small access holes (shown in the foreground of Fig. S2A), which allow access for the perfusion tubing, the glass recording micropipette and the 3 mm O₂ optode (PyroScience GmbH from Ohio Lumex) into the vitreous of the eyecup. While the voltage was measured between the micropipette tip in the vitreous with respect to a ground wire touching the scleral side of the retina, ERGs were evoked by light flashes from the overlying LED. In this configuration, the recorded in vitro ERG of the intact eyecup had all the characteristics of an ERG of the in vivo eye.

Maintaining hypoxia in the special eyecup chamber was difficult owing to the several small holes in the transparent plastic cap that allows for limited and unwanted exposure to ambient air. Therefore, an additional small hole was made in the eye chamber's cap to permit a thin needle to deliver a continuous gas flow of 95% N₂/5% CO₂ to flush out the oxygen contamination from the ambient air. This approach to maintain hypoxia in the retina and limit oxygen contamination was successful to a great extent. ERGs were recorded while the O₂ optode recorded very low oxygen partial pressures in the eyecup. However, in some cases, over time, air would leak into the eyecup chamber and reach the retinal eyecup via the medium bathing the vitreal surface, causing oxygen contamination and slow drifts of increasing P_O2. Note that P_O2 curves did not slowly drift up to 20.0 kPa during hypoxia for preparations 1 or 3, when the brainstem chamber was completely filled with physiological perfusate and covered with a thin glass plate.

All recording micropipettes were pulled from 1.5 mm OD thick-walled borosilicate glass using a horizontal puller (P-87; Sutter Instruments, Novato, CA, USA) and filled with 3 mol l⁻¹ potassium acetate, resulting in 0.5–3 MΩ microelectrodes. The amplified signals were recorded in the current clamp mode of an Axoclamp 2B amplifier and further pre-amplified (Warner Instruments IPF-200; c/o Harvard Bioscience, Holliston, MA, USA) prior to digitization (ClampEx using Digidata). Sensory stimuli (single pulses <20 ms or brief trains of pulses) were recorded either as single traces or as averages of up to 64 traces. At least 3 s separated each stimulus. The ClampEx software sampled those voltages with a 50-µs sampling interval, averaged many dozens of these individual stimuli or pulse trains and then saved those traces as digital files. All digitized traces generated were analyzed offline using Clampfit (Axon Instruments). The waveforms of the field potential responses varied with the micropipette position in the brainstem and were multi-phasic in nature. Moreover, by using an amplifier with only a high-frequency cut-off filter (50 kHz), the baseline value of each trace differed slightly from trace to trace. Therefore, the average value of pre-stimulus component of each voltage trace was digitally subtracted and the post-stimulus waveform was rectified at the computed baseline value. Using these filtered traces, amplitudes were quantified as the areas under the curve in mV×ms for the recordings made from the cochlear nuclei and the OT.

The analysis of ERG responses differed from those recorded in the brainstem because of the two consistent ERG components: an a-wave (hyperpolarization) and a b-wave (depolarization) (Fig. S3B). Six different stimulus intensities were presented in an increasing order of brightness, from near-threshold stimuli to just-saturating stimuli. The ERG traces were quantified using a standard set of formal waveform parameters using the Clampfit software: latency to a-wave onset, latency to a-wave peak, a-wave peak amplitude, a-wave area, b-wave peak amplitude, b-wave area, time between a-wave and b-wave peaks, latency to b-wave peak, maximum slope between a-wave and b-wave, and the time of maximum slope (Fig. S3A).

Single-factor repeated-measures ANOVA was used to determine whether severe hypoxia and re-oxygenation affected the evoked responses of the cochlear nuclei and the OT. For the ERGS, two-factor ANOVA was used to elucidate the effects of both stimulus intensity and oxygenation state, and to determine whether an interaction between the variables existed. Whenever equal variance or normality assumptions were violated, the data were ranked prior to testing (a-wave onset and b-wave implicit times). Statistically significant single factor effects and interactions were further elucidated post hoc with Student–Newman–Keuls testing. All analyses were conducted using Sigmaplot 13 (Systat Software Inc., Chicago, IL, USA). The significance level was always α<0.05.

In the present study, the auditory and visual responses in a reduced brain preparation were examined to evaluate whether the effects of hypoxia on sensory systems would differ depending on the mechanisms that transduce the two environmental stimuli into neural messages within the brain. The first set of experiments studied the mechanoreceptive hair cells of the auditory system and the second set studied the light-sensitive photoreceptors of the visual system.

To determine whether severely hypoxic superfusion affects the mechanosensory functions of the turtle inner ear, voltage recordings were made from the brainstem with a low impedance glass micropipette placed near its connection with the eighth cranial nerve during mechanical pulses of TM (see Materials and Methods, preparation 1 and Fig. 1). A representative response of the evoked potential to severely hypoxic superfusion is depicted in Fig. 2 and mean responses from eight experiments in Fig. 3. The P_O2 of the superfusate decreased to 1.35±0.20 kPa (mean±s.e.m.; range 0.39–2.71 kPa) (Fig. 3A). Despite this, there were no statistically detectable effects of prolonged near anoxia on the responses to TM movements, either for the absolute (Fig. 3B) or normalized (Fig. 3C) response areas. To confirm that we were able to detect a diminished response if it had occurred by another mechanism, we applied lidocaine to the preparation, which strongly attenuated the evoked potential of the TM movements (Fig. 4).

Owing to the lack of an effect of hypoxia on the field potential responses of brainstem neurons produced by mechano-sensitive hair cells in the turtle's inner ear, we performed a set of complementary experiments on the visual system, where the sensory input also originates outside the cranium. In the first set, ERGs were used to measure the retina's visual response as the modulation of the electrical potential at the cornea relative to the sclera. To this end, ERG traces were recorded in response to very brief light flashes from the LED of varying durations (0.05, 0.2, 0.5, 2.5, 5.0 and 10.0 ms) during perfusion with normoxic (control) medium, then with hypoxic medium, and finally, after the return to the normoxic medium (reoxygenation/recovery).

Representative responses of the evoked responses of the retina to the light flashes during hypoxic superfusion are depicted in Fig. 5. The P_O2 of the perfusate decreased to 3.81 and 1.61 kPa (Fig. 5B and E, respectively), producing amplitude decreases by ∼50% in all the peak-to-peak evoked responses during the hypoxia (Fig. 5D–G). There was an overall effect of hypoxia (single factor effect), independent of stimulus intensity on the peak amplitudes of both a- and b-waves, but also an interaction between stimulus intensity and hypoxia. The largest effects were observed at the highest stimulus intensities (0.2–10 ms), whereas the responses at the lowest intensity (0.05 ms) were unaffected (Fig. 6).

The timing of the a- and b-waves was also affected by hypoxia, but in different ways between the two waveforms. The a-wave onset time was significantly longer in the two lowest stimulus intensities (Fig. 7A), whereas the b-wave implicit time showed single-factor effects of hypoxia across all stimulus intensities (Fig. 7B). There was no effect of the hypoxia on a-wave implicit time (data not shown).

The second set of visual response experiments involved the recording of brainstem light responses in the OT that originate in the photosensitive retina while superfusing the preparation with hypoxic medium. Like the averaged field potentials recorded in the auditory brainstem, averaged field potentials were measured as the area under the rectified curve.

An example of a recording of the light-evoked responses of the OT with severely hypoxic superfusate is depicted in Fig. 8. A moderate amplitude response to a weak stimulus (0.5 ms, tan box in Fig. 8C) was chosen to monitor the responses in different media. As described above for the graph in Fig. 2D, the blue line of Fig. 8D shows the quantified responses to evoked rectified field potentials. Examples of color-coded raw traces from this experiment are shown in Fig. 8E, corresponding to the colored symbols in Fig. 8D (see Fig. 8). Within minutes of starting the severely hypoxic perfusion, the P_O2 of the superfusate decreased to less than 1.33 kPa, maximally suppressing the evoked response of the OT within ∼10 min. Upon returning to oxygenated medium, both light responses and P_O2 returned to control levels, also within ∼10 min.

Like Fig. 3A, Fig. 9A depicts overlapping dashed lines of P_O2 optode traces of four of the five OT experiments. In Fig. 9B and C, the absolute and normalized response area values from all five preparations are plotted, respectively. During hypoxic superfusion, the responses significantly decreased to ∼75% of control levels, and then fully recovered upon reoxygenation (P<0.05).

Unlike the short onset latency of brainstem responses of a few milliseconds following TM movement (see Fig. S1), the OT light responses exhibited much slower onset latencies of several dozen milliseconds (Figs S4–S6). It is therefore not surprising that the brainstem responses to trains will respond to individual TM stimuli of interstimulus intervals as short as 2 ms (see also Christensen-Dalsgaard et al., 2012), whereas OT responses to individual flashes within trains blend together, exhibiting temporal response summation even when intervals are as large as 200 ms (Fig. S6).

The onset latencies of OT primary depolarizing waves were also significantly longer than ERG responses, averaging 50 ms for moderate light intensities (Fig. S5A), compared with 14 and 45 ms for the a-wave and b-wave, respectively. Like the ERGs, the onset latency of the OT responses increased as the light stimulus intensity decreased (see Fig. S5A, corresponding to the top trace of Fig. S5D). In contrast, brighter light flashes elicited a second depolarization that grew larger for increasing stimulus intensities. Finally, as shown in Fig. 7B, the implicit time of the ERG b-waves slowed during hypoxia. A similar observation was made of slowed onset latencies for OT responses during hypoxia (Fig. 8E). These effects of timing may not reflect the effects of hypoxia on sensory transduction mechanisms found in the anoxic-tolerant turtle, in that the onset and peak timing of responses to TM pulses and ERG a-waves were not affected by hypoxia in these data presented above. More likely, these are features of neural processing after phototransduction.

This study explored the effects of oxygen deprivation on the sensory brainstem responses recorded in pond turtles, known as one of the most anoxia-tolerant groups of vertebrates. Utilizing an in vitro brain preparation, we demonstrated that the auditory and visual systems exhibit very different responses to hypoxic and near-anoxic conditions. For the auditory system, responses could still be elicited by controlled, periodic movements of the tympanum even when oxygen tension of the perfusion medium fell below 0.93 kPa. In contrast, but not surprisingly, the visual system showed a high sensitivity to reduced oxygen in the perfusate for both the retina and especially the OT. ERG responses were diminished at oxygen tensions as high as 4.0 kPa. We conclude that the turtle's visual system exhibits little to no function under anoxic conditions, and perhaps even during some extended voluntary dives. In contrast, at least at the mechano-transduction level, i.e. the cochlear hair cells, auditory function is maintained.

We showed that field potential responses to TM displacement persisted during severe hypoxic and nearly anoxic conditions in turtles. Is this a common response of the auditory brainstem in all vertebrates, or is it unique to the anoxia-tolerant turtle? Unsurprisingly, auditory function is well studied in mammals as the in vivo auditory brainstem response (ABR), which is measured in humans clinically and in other species in laboratory settings (Attias et al., 1990; Skoe and Tufts, 2018), and is a far-field recording of very small neural events using low impedance surface electrodes on the skin. The averaged waveforms have identifiable voltage trace morphology that reflects different anatomical structures such as the olivo-cochlear bundle, the cochlear nucleus and the inferior colliculus (Kileny et al., 2021). Almost universally, there is a decline in the mammalian ABR in response to severe hypoxia, which contrasts with our observations in turtles. These differences could be attributed to methodological differences. For example, unlike the mammalian ABR, turtle responses were made in vitro and were evoked by direct displacement of the TM and measured using microelectrodes that have small pipette tips within the brainstem. In contrast, mammalian ABR recordings typically rely on a systemic means to lower tissue oxygen levels and use an indirect measure to compute those oxygen levels (Attias et al., 1990; Haupt et al., 1993; Sawada et al., 2001; Sohmer and Freeman, 2001).

Aside from the vast literature on the mammalian auditory response to hypoxia, there are only a few studies in non-mammalian vertebrates, mostly in anoxia-tolerant goldfish. Fay and Ream (1992) recorded spike activity of the afferent fibers from the saccule of anesthetized goldfish in response to a sinusoidal pressure wave produced from an underwater speaker while the gills were either ventilated with aerated water or not at all (asphyxiated). Suzue et al. (1987) conducted a similar experiment in goldfish, but used a speaker in air while ventilating the gills with anoxic water. In both studies, the evoked potentials were diminished by anoxic ventilation or asphyxia, indicating that the auditory response of the goldfish is sensitive to oxygen deprivation, unlike the turtles in the present study.

Although pond turtles are uniquely adapted to both terrestrial and aquatic life, hypoxic environments are predominant in eutrophic habitats such as muddy bogs, where pond turtles frequently live. In an experiment that studied auditory thresholds of turtles, either submerged or airborne, head withdrawal behaviors were measured in response to sounds paired with an aversive stimulus (Patterson, 1966). This behavioral response showed that turtles were more sensitive to underwater stimuli than to airborne stimuli. Similarly, ABR was measured in airborne and underwater turtles and their sound thresholds in water were 20–30 dB lower than those measured for airborne sounds (Christensen-Dalsgaard et al., 2012). Taken together, a pond turtle's hearing is very likely to be effective even in hypoxic water owing to increased auditory sensitivity in that environment.

It is important to note that our findings are limited to more peripheral stages of the auditory system, where: (1) the otic capsule and temporal bone were both bone ischemic, eliminating possible humoral influences, (2) the production of healthy endolymph necessary for hair cell mechanotransduction was functional even when bathed in hypoxic medium and (3) hypoxic hair cells maintained synaptic communication to ganglion cells and still relayed action potentials along axons to the brainstem. Analyses of auditory responses of intact turtles would be useful to understand the central auditory effects of hypoxia (Christensen-Dalsgaard et al., 2012; Patterson, 1966; Piniak et al., 2016; Willis, 2016).

The amplitudes of both ERGs and OT field potentials were decreased when superfused with severely hypoxic medium, a predictable outcome given that phototransduction is mediated by an energetically expensive enzymatic cascade found in the retinal photoreceptors and previous published ERG data from anoxic turtles (Stensløkken et al., 2008). The retina in general and the photoreceptors in particular use the highest amount of metabolic energy of all vertebrate sensory systems (Country, 2017).

Experiments of retinal cellular physiology indicate that P_O2 in the outer retina is most modulated by retinal illumination at the photoreceptor layer (Linsenmeier, 1986), more so in cones as compared with rods (Ingram et al., 2020). Presumably, the highest oxygen need would be in animals with cone-dominated retinae. The aquatic turtle should therefore be very sensitive to retinal anoxia, especially considering its cone-dominated retina and its lack of a blood supply to the outer retina (Damsgaard et al., 2019).

The present study of anoxic turtles was consistent with an earlier ERG study of anesthetized, nitrogen-ventilated turtles (Stensløkken et al., 2008). Although those in vivo ERGs were recorded using 10-fold longer light flashes during 2 h of anoxia, those ERGs were qualitatively similar to those of our preparation in that the b-waves in both studies were significantly diminished by the anoxia. In vivo a-waves were significantly decreased, but their responses were smaller, like the a-waves in our study that tended to decrease, but also did not achieve statistical significance. Other reports found that a-wave responses in both hypoxia-intolerant and hypoxia-tolerant vertebrates are less sensitive to hypoxia than b-wave responses and central visual responses (Johansson et al., 1997; Niemeyer, 1975; Tinjust et al., 2002). From these comparisons, we conclude that the responses to anoxia of the b-wave observed in vivo are intrinsic to the eyecup, and may likely reflect the phototransduction cascade that occurs within photoreceptor outer segments.

Compared with hypoxic changes measured in the retina, those effects recorded in the OT were quantitatively greater. This difference might be explained simply if hypoxia effects within the cranium occur independent from the response depression within the retina, and both ocular and midbrain effects are additive.

Another explanation for the larger hypoxic effects in the OT as compared with the retina is that the experimental recordings of the tectal structure differ both in the different anatomical forms and the methods of response quantification. ERGs represent averaged, low-frequency trans-retinal currents, as compared with the tectal field potentials, derived from multi-unit spiking neurons. Those responses were also quantified using different physiological measures. ERGs are summed voltage changes across the entire retinal surface, whereas the OT electrode just received input from a localized sphere of neural tissue that maps a small area of visual space onto the tectal surface. Moreover, the neural architectures of these two visual structures differ substantially: retinal neurons relay graded potentials along the narrow radial retinal path, whereas tectal neurons gather synaptic inputs from large, complex dendritic trees that project laterally across the entire tectal surface.

A common feature of ERGs and OT field potentials is that their light responses were decreased during superfusion with anoxic medium. Light responses during hypoxia are also decreased in mammals including humans. Using isolated rabbit retina in vitro, complete anoxia blocked all retinal light responses (Ames and Gurian, 1963). Using intact mammals, severe hypoxia or total anoxia is not feasible (but see Eysel, 1978). However, mildly hypoxic cats showed differential effects on the a-wave versus b-wave response, depending on the level of hypoxia (Derwent and Linsenmeier, 2000; see also Noell and Chinn, 1950). Low hypoxia (P_O2 of 6.7–8.0 kPa) reduced the a-wave amplitude by small amounts, whereas an 8.9% response decrease occurred during more severe hypoxia (2.7–4.0 kPa). The same severe hypoxia had a strong (35%) effect on the b-wave amplitude. Humans breathing mildly hypoxic air (12% O₂/88% N₂) exhibited a statistically significant decrease in their b-wave amplitude but no change in their a-wave amplitude (Tinjust et al., 2002). Therefore, like the turtle ERG, the mammalian a-waves derived from photoreceptor activity are more resistant to hypoxia than b-wave activity of the inner retina.

The ability of the anoxia-tolerant crucian carp to respond to visual stimuli has also been investigated during anoxia. Unlike pond turtles, crucian carp are not quiescent when exposed to anoxic conditions (Nilsson et al., 1993), which suggests that their visual system may be better able to respond to light during anoxia, unlike pond turtles. Yet, ERGs and OT field potentials of the crucian carp showed a strong decrease in visual responses (Johansson et al., 1997). However, the rate of the decrease of the carp's light response amplitudes were slower than the decrease rates observed in the turtles, perhaps related to the unique fish eye (Country, 2017) or the ability of the in vitro turtle preparation to reach anoxic levels quickly by direct perfusion.

The present study found that visual responses were significantly decreased in both the retina and the OT during hypoxic conditions (P_O2<4.0 kPa), whereas responses to auditory stimuli (direct vibration of the TM) remained in greater hypoxic conditions (P_O2<2.7 kPa). The physiological mechanisms that underlie a turtle's auditory capacity to tolerate hypoxia/anoxia may provide valuable insights into an organism that is subjected to anoxia, specifically, how animals manage with decreased ATP availability. Their survival usually requires organ-specific metabolic suppression (Hochachka, 1986) and physiological functions are likely prioritized, with less essential and more energetically expensive ones suppressed, and more essential and energetically inexpensive ones preserved (Jackson, 1987). Sensory functions pose a particularly compelling problem because their relatively high energetic cost will affect the animal's survival. Studying sensory function in anoxia-tolerant animals such as turtles may enable us to understand where the different sensory functions sit on this spectrum, while also revealing fundamental mechanisms of sensory function in response to low oxygen.

The intact pond turtle is a model organism that can be observed recovering from severe hypoxia after extensive periods of time. That recovery requires overwintering in near-zero temperatures. Freshwater turtles that spend their winters at those temperatures enter the nearby ponds to survive. Survival in the resulting anaerobic environment requires many biological adaptations that are only successful in near-freezing temperatures. In contrast, these in vitro experiments were performed by controlling the P_O2 directly in the physiological medium without reducing temperature of that medium. Obviously, the survival of an intact turtle in its natural environment may also be dependent upon other factors: chemistry of the blood and other systemic circulating modulators, pH of the cerebrospinal fluid, and the durations of acute temperature exposure and/or temperature acclimation. Future in vitro studies should evaluate the effects of lower temperatures and other systemic parameters in normoxic medium in combination with hypoxic medium.

The ERG studies were performed in partial fulfillment of a Master's thesis (S.A.).