State-Dependent Changes in Cortical Gain Control as Measured by Auditory Evoked Responses to Varying Intensity Stimuli

Abstract

Study Objectives:

Auditory evoked potential (AEP) components correspond to sequential activation of brain structures within the auditory pathway and reveal neural activity during sensory processing. To investigate state-dependent modulation of stimulus intensity response profiles within different brain structures, we assessed AEP components across both stimulus intensity and state.

Design:

We implanted adult female Sprague-Dawley rats (N = 6) with electrodes to measure EEG, EKG, and EMG. Intermittent auditory stimuli (6-12 s) varying from 50 to 75 dBa were delivered over a 24-h period. Data were parsed into 2-s epochs and scored for wake/sleep state.

Results:

All AEP components increased in amplitude with increased stimulus intensity during wake. During quiet sleep, however, only the early latency response (ELR) showed this relationship, while the middle latency response (MLR) increased at the highest 75 dBa intensity, and the late latency response (LLR) showed no significant change across the stimulus intensities tested. During rapid eye movement sleep (REM), both ELR and LLR increased, similar to wake, but MLR was severely attenuated.

Conclusions:

Stimulation intensity and the corresponding AEP response profile were dependent on both brain structure and sleep state. Lower brain structures maintained stimulus intensity and neural response relationships during sleep. This relationship was not observed in the cortex, implying state-dependent modification of stimulus intensity coding. Since AEP amplitude is not modulated by stimulus intensity during sleep, differences between paired 75/50 dBa stimuli could be used to determine state better than individual intensities.

Citation:

Phillips DJ; Schei JL; Meighan PC; Rector DM. State-dependent changes in cortical gain control as measured by auditory evoked responses to varying intensity stimuli. SLEEP 2011;34(11):1527-1537.

INTRODUCTION

Sleep significantly alters cortical responses to sensory stimulation^1–5; however, little is known about the mechanisms underlying these effects. A better understanding about the input/output relationships of stimulus intensity modulated responses at different sleep stages within the sensory pathway could provide insight into processing of sensory input across sleep state, assist in understanding the control and function of sleep, and lead to better metrics for determining sleep and potential localized sleep states.^6,7 To investigate the relationship between stimulus intensity and sensory response amplitude for different structures across different states, we implanted rats with cortical electroencephalogram (EEG) electrodes and provided different auditory-stimulus intensities across 24-h recording periods. Since various auditory evoked potential (AEP) components correspond to different brain structures within the sensory pathway, these components could also provide signatures for state-dependent regulation of structures responsible for processing external stimuli.

To understand potential sleep effects on auditory AEPs, the pathway for auditory stimuli from the ear to the cortex can be divided across 6 main structures (Figure 1). At the level of the cochlea and auditory nerve, sound-pressure waves produce action potentials at a rate modulated by intensity,^8,9 creating a compound action potential within the auditory nerve. The earliest AEP components may arise from activity that ultimately originates at this level.^10–13 The cochlear nucleus projects to the superior olive, then through the lateral lemniscus and inferior colliculus, with another synapse in the medial geniculate body of the thalamus (MGB), arriving at the primary auditory cortex. Secondary cortical structures produce later AEP components in response to complex and meaningful stimuli; however, because we used simple click stimuli, the later components were not the focus of this study. Thus, each step in the auditory pathway generates electrical signatures of neural activity which are present in the AEP. Sleep regulatory mechanisms could, in principle, modify the activity at any and all structures involved in auditory processing. Evaluation of specific AEP temporal characteristics could lead to better tools for determining animal vigilance state in addition to tissue functional state at various levels within the brain.

Figure 1.

The auditory pathway consists of 6 main structures that contribute to various components of the auditory evoked response potential. Once the sound pressure waves are transduced by the cochlea, action potentials travel through the auditory nerve (AN) to the cochlear nucleus (CN), which then projects to the superior olive (SO). From the superior olive, projections through the lateral lemniscus (LL) go to the inferior colliculus (IC), and then on to the medial geniculate body of the thalamus (MGB), finally arriving at the primary auditory cortex (Au1).

We illustrate our rodent AEP definitions from a typical trace shown in Figure 2. To focus our analysis on specific temporal events that occur consistently across animals, we limited our analysis to the 3 most robust and significant components of our recordings. One of the earliest components, defined as the early latency response (ELR, Figure 2C), had a peak that occurred roughly 2-3 ms after the stimulus and a trough at 4-5 ms. This response could reflect activation of neural generators within the brainstem leading into the inferior colliculus.^11–13 The middle latency response peak (MLR, Figure 2B) typically occurred 8-10 ms after the stimulus, while the trough occurred 12-15 ms post-stimulus. The MLR response could be generated by the MGB, and primary cortical structures within layer 4.^14,15 The MLR leads into the late latency response peak (LLR, Figure 2A) that appeared 25-30 ms after the stimulus, with the trough at 45-100 ms. As higher order processing occurs, it is difficult to identify the neural generators for specific waveform patterns. The MGB, hippocampus, primary and secondary auditory cortex, and the reticular activating system have all been suggested as possible contributors to the LLR.^3,13,16

Figure 2.

A representative average auditory evoked response potential from a rat illustrates the various components that can be detected at the cortical surface. The top panel (A) shows an averaged response to 75 dBa stimuli during wake. The late latency response (LLR) peak and trough, which we used for subsequent analysis, are marked with arrows. Other late components are also present; however, with longer latency peaks it become increasingly difficult to determine the neural structures of origin. The middle panel (B) shows an expanded view of the first 25 ms of the response, showing the middle latency response (MLR) peak and trough. The lower panel (C) is further expanded in time to show the first 6 ms after the stimulus. Several early peaks can be identified which may originate from population action potentials generated by the cochlea (see text). The most consistent early latency response (ELR) peak and trough are marked with arrows, and were used in further analysis of the early component.

Several studies have shown that sleep state alters sensory information along the auditory pathway for both field evoked response potentials and unit activity. The cochlear nucleus,¹⁷ lateral superior olive,¹⁸ inferior colliculus,¹⁹ thalamus,²⁰ and auditory cortex²¹ all exhibit state-dependent modulation of evoked firing rates to auditory stimuli. The presence of spindles can significantly modify AEP amplitude across moderate and very loud intensity stimuli,²² and close examination of AEPs across behavioral state shows that cortical AEPs are still present and can be larger during sleep compared to wake and REM sleep.^2–5,23 Even more profound effects are produced by different anesthetics.^24–31 However, during waking, an animal is able to provide appropriate responses to its environment, while responsiveness to external stimuli is reduced or absent during sleep. One main difference between sleep and anesthesia is that unconsciousness can be reversed from sleep in order to react to salient and relevant stimuli³²; thus a certain amount of sensory processing must occur during sleep. Many of these studies have focused on a specific component of the AEP and used a single intensity auditory stimulus to identify differences across sleep and wake. Our current study offers a comprehensive analysis of the 3 most robust and significant components of the AEP, and identifies changes to varying intensity across sleep and wake states.

Since synaptic activity can control the gain and sensitivity of cortical neurons,³³ the present study may help shed light on the mechanisms responsible for AEP differences between sleep and wake. Two questions in particular need to be addressed. First, what structures along the sensory pathway are subject to change in amplitude with changes in state? Tracking the various temporal components of the AEP across state may correspond to the responses of particular structures.³⁴ Second, what mechanisms are responsible for those AEP components that are sleep dependent? Such mechanisms include the possibility that different numbers of cells may be recruited, activation may occur more synchronously during sleep, thus temporal summation may differ on the cortical surface, or they may depend on membrane potentials that fluctuate between hyperpolarized (down) and depolarized (up) states during sleep.^35–38 If the activation profile of cortical neurons depends on baseline membrane potential and conductance,³³ then we would expect AEPs to also depend on membrane properties and appear different during sleep due to these fluctuations.^7,39

In summary, state-specific AEP changes may provide insight into state-dependent cortical activity and provide predictive measures of sleep state. Since increasing evidence supports the notion that sleep can occur in localized brain regions as small as the cortical column,^6,7 AEP signatures of local sleep may also provide predictive measures of performance during critical tasks.⁷ However, before AEPs can be used as a reliable probe, we must first identify appropriate stimulus parameters and understand how they may influence specific AEP components. We initially hypothesized that all AEP components would follow a typical sigmoid curve relationship increasing in amplitude across auditory stimulus intensity with the slope of this relationship altered by state-specific modulation on the structures generating each component. However, our findings showed that this relationship did not hold true for the late AEP components during quiet sleep.

METHODS

Experimental Procedures

Surgical procedures

Adult female Sprague-Dawley rats (N = 6, 240-300 g, Simonsen Laboratories, Gilroy, CA) were anesthetized using 2.5% isoflurane and chronically implanted with electroencephalogram (EEG), neck electromyogram (EMG), and electrocardiogram (EKG) electrodes. The dorsal skull surface was exposed and stainless steel screw electrodes (J.I. Morris, Southbridge, MA, F00CE188) were placed over the frontal (1 mm rostral to bregma and 2 mm lateral to midline) and parietal (1 mm lateral to the temporal ridge and 6 mm caudal to bregma) lobes, with 2 ground references over the cerebellum (1 mm caudal to lambda and 2 mm lateral to midline). A multi-stranded Teflon-insulated stainless steel wire (New England Wire, Lisbon, NH, 212-50F-357-0), with 2 mm of exposed wire was inserted into the neck muscles to record neck EMG. Another wire with 4 mm exposed was inserted subcutaneously along the thoracic cavity to record EKG and thoracic EMG. All electrodes were connected to a plug adapter used to interface with the data system. Additional screws were placed over the parietal/temporal lobes, 4 mm rostral to lambda, and lateral to the temporal ridge to assist in anchoring. Dental acrylic enclosed the electrode wires and sealed the surgical site, leaving only the connector exposed. Broad-spectrum antibiotic ointment was applied to the skin around the head stage. To prevent pain and infection, analgesic Flunixin Meglumine (1.1 mg/kg) and antibiotic Gentamicin (5 mg/kg) were injected subcutaneously up to 3 days following surgery. The Washington State University Animal Care and Use Committee reviewed and approved all surgical procedures.

Recording procedures

Recording chambers were 30×30×50 cm acrylic enclosures and animals were tethered through a thin cable/commutator setup (ProMed-Tec, Bellingham, MA, Pro-ES24) to the data acquisition system. Animals were adapted to the recording chamber during ≥ 8 sessions of 30 min each, with stimuli at 65 dBa present. Lighting was maintained at a 12-h on/off cycle, and food and water were available ad libitum during the 24-h recording session. Electrophysiological data were filtered between 0.1 Hz and 3.2 kHz and digitized at 10-kHz sample rate using a custom data system that allowed continuous data display and storage.⁴⁰ A webcam equipped with infrared light was also used to record the animal's behavior at 1 frame per second.

Stimulus Paradigm

Wide band speaker clicks of varying intensity were generated by manipulating the drive voltage delivered to a speaker with a 0.2-ms square wave pulse. Stimuli were presented at infrequent and random intervals under several experimental paradigms, including: single intensity stimuli (50, 55, 60, 65, 70, or 75 dBa, randomized 6- to 12-s inter-stimulus interval [ISI]), paired stimuli (50 dBa followed 1 s later by 75 dBa, or 75 dBa followed 1 s later by 50 dBa; each pair was randomized between 3-6 s ISI), or 2 intensities presented randomly with equal probability (50 and 75 dBa 3-6 s, randomized ISI) (Figure 3). With the paired presentation, a 1-s delay was chosen to minimize the effects of sensory gating (depression) on paired pulse stimulation.⁴¹ Stimulus intensity ranged from 50 to 75 decibels A-weighted sound pressure level (dBa) in 5 dBa increments calibrated with a sound meter placed in the enclosure (Model 732A, BK Precision, Yoruba Linda, CA). Total and normalized stimulus presentation data is shown in Table 1. One-way ANOVA procedures were used to determine any statistical differences between intensities within a state.

Figure 3.

Examples of our experimental design and stimulus presentation. A schematic example of our experimental protocol (A). Animals were first acclimated to the recording setup during ≥ 8 half hour sessions with 65 dBa stimulation. After the acclimation, the 7 intensities were randomized and recorded individually over a 24-h period. After the single-intensity recordings, the paired and unpaired protocols were randomized and recorded over 24 h; 60-sec examples of our stimulus paradigm are shown for the single intensity (B) 6-12 s randomized, the 75-50 dBa paired stimuli (C) 3-6 s randomized, and 50 and 75 dBa unpaired (D) 3-6 s randomized.

Table 1.

Distribution of stimuli

Total Stimuli	50 dB	55 dB	60 dB	65 dB	70 dB	75 dB
Wake	4884 ± 1039	4386 ± 781	3948 ± 806	4848 ± 558	4674 ± 966	4508 ± 899
LS	1303 ± 442	1334 ± 288	1259 ± 187	1715 ± 262	1420 ± 284	1662 ± 504
DS	1445 ± 147	1320 ± 245	1260 ± 228	1413 ± 267	1337 ± 308	1294 ± 408
REM	414 ± 86	405 ± 84	413 ± 45	454 ± 65	497 ± 93	392 ± 139
Normalized Stimuli	50 dB	55 dB	60 dB	65 dB	70 dB	75 dB
Wake	0.61 ± 0.02	0.60 ± 0.02	0.56 ± 0.04	0.58 ± 0.03	0.59 ± 0.02	0.60 ± 0.05
LS	0.15 ± 0.03	0.17 ± 0.01	0.19 ± 0.02	0.20 ± 0.01	0.18 ± 0.00	0.20 ± 0.03
DS	0.19 ± 0.03	0.17 ± 0.01	0.19 ± 0.01	0.16 ± 0.02	0.17 ± 0.02	0.16 ± 0.03
REM	0.05 ± 0.01	0.05 ± 0.01	0.07 ± 0.01	0.05 ± 0.00	0.06 ± 0.00	0.05 ± 0.01

Total stimuli (top) and normalized total number of stimuli (bottom) ± SEM across all animals presented between wake, light sleep (LS), deep sleep (DS), and REM for each intensity. There were no statistical differences in the number of stimuli presented in each state across intensity. ANOVA: Wake (P = 0.88) LS (P = 0.58) DS (P = 0.86) or REM (P = 0.61).

Data Analysis

Archived data were parsed into 2-s epochs and converted into their frequency domain by fast Fourier transform (FFT). Data were then divided into 4 states defined by their EMG and EEG power. Our analysis focused on wake (high EMG and low delta power), light quiet sleep (LS, low EMG, medium delta power), deep quiet sleep (DS, very low EMG, high delta power), and rapid eye movement (REM) sleep (flat EMG, high theta, and low delta power). Sleep state was first determined by plotting EEG delta power vs EMG power and using a cluster cutting technique,⁴² then visually confirmed by reviewing the physiological traces and adjusted as needed. To simplify analysis of the data from paired stimuli and 2 random intensity stimuli, LS and DS periods were pooled, and we refer to this state as quiet sleep (QS).

For each stimulus, the corresponding AEP recorded between the parietal electrode and cerebellum reference was sorted by the sleep state during which it occurred, and by stimulus intensity (Table 1). AEPs were averaged across state and stimulus intensity, then the 3 components (as defined in Figure 2) were measured from peak to trough. To control for AEP variations across animal, each amplitude and latency measurement was normalized to the waking 50 dBa value. Each component's relative latency and amplitude was compared and tested for significant differences from 50 dBa within a state and for differences from wake across state at the same stimulus intensity using a Mann-Whitney U-test to assess the potential for paired pulse depression.⁴¹ When stimuli were paired (50 dBa followed 1 s later by 75 dBa, or vice versa), we measured the LLRs for each intensity within the pair, then compared the difference in amplitude using Mann-Whitney U-tests. A nonparametric test was used because a Gaussian distribution could not be assumed from our data set. To track changes in single trial AEPs across time, individual cortical response amplitudes were calculated using a curve fitting routine that fit individual AEP responses to the averaged response.⁴² The amplitude difference between each 50 dBa and 75 dBa pair was calculated, plotted across time and compared to the sleep hypnogram for the same period. The 50 and 75 dBa amplitude difference plots were smoothed using a 100-point boxcar algorithm to compare the moment-by-moment correspondence to sleep state.

RESULTS

Early Latency Response

Averaged AEPs exhibited increased ELR amplitude with increasing stimulus intensity for all sleep states (Figure 4, ELR peak [EP] to ELR Trough [ET]). When intensity was compared within state, significant ELR amplitude increases from 50 dBa were observed at 65-75 dBa during wake, at 60-75 dBa during REM (P < 0.05), and at 55-75 dBa during light quiet sleep (LS) and deep quiet sleep (DS) (^#P < 0.05) (Figure 5A). When same-intensity stimuli were compared to wake, ELR amplitude from LS, DS, and REM were significantly larger at intensities 50-60 dBa (*P < 0.05). However, as stimulus intensity increased to 65-75 dBa, ELR amplitudes during QS were not different from wake (Figure 5A). ELR peak and trough latencies did not show significant differences across state or intensity (Figures 5B and 5C).

Figure 4.

Example traces of the first 20 ms after the stimulus from one animal across all stimulation intensities and sleep states show the relative magnitudes of the various early components. Stimulus time is marked with vertical gray lines. The early latency response (ELR) peak (EP) and trough (ET) can be seen in all traces. The ELR increases with stimulation intensity across all states. The middle latency response (MLR) peak (MP) and trough (MT) also increases in amplitude during wake, but less consistently during light sleep (LS) and deep sleep (DS). During rapid eye movement sleep (REM), the MLR response is attenuated at all stimulation intensities.

Figure 5.

To compare evoked response components across animals, all measurements were normalized to the waking response, 50 dBa intensity stimulation. The top panel (A) plots the average early latency response (ELR) amplitude for all animals across 6 stimulation intensities and 4 sleep states including wake, light sleep (LS), deep sleep (DS), and rapid eye movement sleep (REM). The ELR amplitude increased with louder stimulus intensities across all states. The lower intensity stimuli produced higher ELR amplitudes during all sleep states compared to wake. Significant differences across stimulation intensity compared to the 50 dBa level within a state are marked (^#P < 0.05 Mann-Whitney U test). Significant differences across state, compared to wake at the same intensity level are marked (*P < 0.05 Mann-Whitney U test). Panels (B) and (C) show changes in the ELR peak and trough latencies, respectively. No significant differences in latency across state or intensity were found.

Middle Latency Responses

The MLR components shown in Figure 4 (MLR peak [MP] to MLR trough [MT]) revealed large amplitude increases with higher stimulus intensities during wake, especially at 75 dBa (^#P < 0.05) (Figure 6A). The variability in the MLR during LS and DS was significantly greater, with a trend for the amplitudes to be augmented relative to the wake responses, yet with less dependence on stimulus intensity between 50 and 70 dBa (Figure 6A). The MLR was significantly larger at 70-75 dBa for wake and at 75 dBa during, LS and DS (^#P < 0.05), but no statistical increase was observed during REM. The MLR during REM sleep appeared to be attenuated or nonexistent for intensities > 65 dBa relative to wake. No significant difference in MLR amplitude was found during REM sleep over the intensity range tested. When compared to the wake state, MLR amplitude was significantly smaller at the 70-75 dBa intensities during REM (*P < 0.05) (Figure 6A). As with the ELR, no significant differences in MLR peak or trough latency were observed across state or intensity (Figures 6B and 6C).

Figure 6.

The average middle latency response (MLR) amplitudes for all animals across state and stimulus intensity were normalized to the average 50 dBa wake value. During wake, increased stimulation intensity elicited larger MLR amplitude components. A similar trend can be seen during light sleep (LS) and deep sleep (DS); however, this did not show significance due to the high variability. During rapid eye movement sleep (REM), the MLR stayed at a very low level, and was severely attenuated at the high stimulus intensities. Significant differences across stimulation intensity compared to the 50 dBa level within each state are marked with a (^#P < 0.05 Mann-Whitney U test). Significant differences across state, compared to wake, at the same intensity level are marked with a (*P < 0.05 Mann-Whitney U test). Panels (B) and (C) show changes in the MLR peak and trough latencies respectively. No significant differences in latency across state or intensity were found, though there was a trend for shorter latencies with increasing stimulus intensity for wake, LS, and DS.

Late Latency Responses

During wake and REM, LLR amplitude increased with stimulus intensity (Figure 7, (LLR peak [LP] to LLR trough [LT]) with 60-75 dBa stimuli generating significantly higher LLR responses than 50 dBa stimuli (^#P < 0.05) (Figure 8A). However, during LS and DS, the LLR amplitude did not change with stimulus intensity (Figure 8A). At 50-60 dBa, the response during LS and DS was significantly larger than those from the wake state (*P < 0.05). At 65-70 dBa, the LS and DS LLR amplitudes were not different from wake. When stimulus intensity was increased to 75 dBa, the LLR amplitude during LS, DS, and REM was significantly smaller than the wake response (Figure 8A) (*P < 0.05). While the LLR peak and trough latencies did not significantly change within a state, LLR peak latency was significantly longer for every stimulus intensity during REM when compared to wake (Figure 8B) (*P < 0.05), and LS, DS, and REM trough latencies were all significantly longer when compared wake (Figure 8C) (*P < 0.05). When amplitude data across stimulus intensity were plotted in line graph form to reveal stimulus/response profiles, sigmoidal-like curves are present for ELR and MLR components (except MLR during REM sleep) and LLR during wake, but were absent from the LLR during QS and are muted during REM (Figure 9). Additionally, there were no significant differences between stimuli presented as a single intensity vs the unpaired randomized 2-intensity paradigms (P = 0.56), and the data were pooled for this analysis.

Figure 7.

Example traces from 150 ms after the stimulus from one animal show the state and stimulus intensity changes in the late latency response (LLR). The LLR peak (LP) and trough (LT) amplitudes increased consistently with increased stimulation intensity during wake and rapid eye movement sleep (REM), but were at a constant level during light sleep (LS) and deep sleep (DS). We also observed a significantly longer LLR trough latency during LS and DS due to the presence of a second waveform that extended the duration of the LLR.

Figure 8.

As with Figures 5 and 6, the average late latency response (LLR) amplitudes for all animals across state and stimulus intensity were normalized to the average 50 dBa wake value. During wake and rapid eye movement sleep (REM), increased stimulation intensity elicited larger LLR amplitude components. No significant changes in LLR amplitude were seen during light sleep (LS) or deep sleep (DS). This result caused the low level stimuli (50 to 60 dBa) to appear larger during LS and DS than wake, and high level stimuli (70 to 75 dBa) to appear smaller. During REM, the highest stimulus intensity LLR was significantly lower when compared to the wake value. Significant differences across stimulation intensity compared to the 50 dBa level within a state are marked with a (^#P < 0.05 Mann-Whitney U test). Significant differences across state compared to wake, at the same intensity level are marked with a (*P < 0.05 Mann-Whitney U test). Panels (B) and (C) show changes in the LLR peak and trough latencies respectively. No significant effect of stimulation intensity was found on peak and trough latency. As with the MLR latencies in Figure 6, there was a trend during wake, LS, and DS for latencies to decrease with increased stimulus intensity. The LLR peak latency was significantly longer during REM when compared to wake, and all 3 sleep states exhibited increase LLR trough latency compared to wake.

Figure 9.

Line plot representations of the ELR, MLR, and LLR amplitude data described in Figures 4, 5, and 7. Response curves are present for the ELR, MLR (except MLR during REM), and LLR during wake, but are absent during QS and appear muted during REM. Significant differences across stimulation intensity, compared to the 50 dBa (^#P < 0.05 Mann-Whitney U test) and across state, compared to wake, (*P < 0.05 Mann-Whitney U test) were identified.

Paired Stimulus Paradigm

When 2 stimuli with different intensities were paired during a recording (50 dBa followed 1 s later by 75 dBa, 75 dBa followed 1 s later by 50 dBa, or 50 dBa and 75 dBa randomized), the same average response component relationships between stimulus intensity and state for the LLR was observed as described for single stimuli (Figure 10) (P = 0.57). Thus, the effect of pairing the stimuli with a 1-s delay did not significantly change the state or intensity characteristics of the responses as would be expected from the literature.⁴¹ The LLR amplitude was significantly higher at 75 dBa stimulus intensity than 50 dBa during wake, and responses during LS and DS showed no significant change in amplitude between the 2 stimulus intensities. We then calculated the LLR amplitude difference between the paired stimuli and plotted each LLR amplitude difference over time, superimposed with the state hypnogram (Figure 11). The smoothed data show > 100 μV difference in LLR amplitude during wake and REM sleep, and smaller differences (0-25 μV) during quiet sleep (QS, which includes both LS and DS).

Figure 10.

Auditory evoked potentials (AEP) from wake, sleep, and rapid eye movement sleep (REM) when stimuli were paired (50 dBa followed 1 s later by 75 dBa and vice versa) show the same relationship in amplitude to the single-intensity trials. In both paradigms the late latency response (LLR) to 75 dBa stimuli (black line) during wake and REM were significantly larger than LLR to 50 dBa stimuli (gray line). LLR amplitude during LS and DS also follow their single-intensity relationship, and do not significantly vary between the 50 and 75 dBa pair stimuli.

Figure 11.

When 75 dBa and 50 dBa stimuli were paired with a 1-s delay, the same relationship in the late latency response (LLR) amplitudes were observed as was described for single stimuli in Figures 7 and 8. However, with the paired stimulation, we plotted the LLR difference for each stimulus event across time, superimposed on the animal sleep state as scored by the EEG and EMG parameters. In this figure, a 24-h period is plotted from 2 animals to illustrate the LLR amplitude differences across time. A blue background represents a quiet sleep period, an orange background represents a rapid eye movement sleep (REM) sleep period, and a white background represents wake. LLR amplitude differences between 75 and 50 dBa pairs are shown in the thick black trace with the mean amplitude represented by the thin horizontal black line. Since it is difficult to accurately estimate the LLR amplitude from individual responses due to ongoing EEG activity (see text), we smoothed the LLR difference plot with a 100-point moving boxcar filter. The smoothed data show that whenever the animal enters quiet sleep, the difference is usually below the mean, but usually increases during waking and REM sleep.

DISCUSSION

Consistent with our previous studies, the LLR component of the AEP during both LS and DS was larger when compared to wake when using the same intensity (50 to 60 dBa) auditory stimuli.^5,42 However, as stimulus intensity increased, the wake LLR amplitude also increased, while the LLR during LS and DS stayed roughly the same. Thus, at 65-70 dBa stimulation, no differences in AEP amplitude were found. At 75 dBa, the LLR component during wake reversed its relationship and was significantly larger than its LS and DS counterparts. This result may clarify earlier results found by some investigators who reported smaller responses during sleep compared to wake,⁴³ while others reported larger responses during sleep,^3,5 and underscores the importance of considering the stimulus intensity when comparing AEP differences across sleep and wake. With high stimulus intensities, the AEP during waking could be larger than that during QS because the QS response does not appear to change across stimulus intensities. Note, however, that within a particular period, AEP metrics can show changes that do not correspond to whole animal state, indicating a brief transition, undetectable by EEG measures, or localized appearances of a different state within a predefined whole animal state.

Since animals are not aroused from sleep as a result of auditory stimuli between 50 and 75 dBa,⁴⁴ this range of stimulation intensities may provide a good probe for AEP characteristics across state. Ideally, a more complete stimulus/response curve (Figure 9) could be generated with higher stimulus intensities. Unfortunately, we could not test stimulus intensities above 75 dBa due to the potential consequences of arousal on the AEP late components.

Since the cortical (LLR) component of the AEP remains constant over different stimulation intensities during QS, the hypotheses that more cells are recruited during the QS response or cells exhibit increased synchrony during QS are not supported. If either of these mechanisms occurred, then we would expect the LLR component to maintain its stimulus intensity/response amplitude relationship. Cortical neurons in their down state might require a higher stimulation intensity to reach the threshold for AEP generation.³³ Since our lowest intensity was 50 dB, this stimulus level may have been high enough to elicit a response in the down state. The present results show that cortical structures responsible for the later AEP components produce constant amplitude responses to the different external stimulus intensities during QS. Since rats can hear sounds at much lower intensities (< 40 dBa), if we were able to assess responses from lower intensities, we may have found an intensity which would elicit a response during wake (or up state), but not during the down state.⁴⁵

Since QS is characterized by fluctuations of cortical neurons between hyperpolarized and depolarized states at a delta rhythm, it is possible that the gain and sensitivity of cells change during QS could lead to the AEP results we observe. Membrane conductance during the up state is relatively high, increasing responsiveness to lower intensity stimulation. The up state is also characterized by a decrease in slope associated with the input/output relationship. During the down state, membrane conductance is low, requiring an increase in stimulation intensity to reach threshold. The input/output relationship during the down state is significantly right-shifted with a steep slope.³³ Thus, with different stimulation intensities, we would expect higher consistency during the down state leading to consistent AEPs across stimulus intensity, whereas during the up state, also typical of waking, the slope is more gradual, leading to a graded response. The presence of the up state during QS delta rhythm may, in fact, dilute the constant amplitude response during sleep; thus if we were able to separate the different phases of the delta rhythm during sleep, differences between up and the down state during QS may be enhanced. The AEP during REM sleep consistently resembled those observed during the wake state. This result suggests that responses during REM sleep may correspond with the up state.

As a consequence, the cortical AEP may also be processed differently during sleep compared to wake. While a response is certainly generated during both states, the external stimulus may be perceived and attended to during wake, while the response is apparently not perceived, nor attended to during QS. In spite of the lack of LLR modulation during sleep, some forms of learning and perception can still occur during sleep. For example, arousal from sleep can correspond to salient or relevant stimuli,³² and classical conditioning can be acquired during sleep,¹ presumably through a certain amount of higher level AEP processing that remains intact.⁴⁶

The early AEP components exhibited stimulus intensity evoked response relationships that were mostly preserved across sleep state. This result indicates that lower structures within the brainstem and midbrain may not be affected by sleep in the same way as higher structures. The ELR amplitude and latency relationships were nearly identical across state, suggesting that sleep did not greatly influence auditory signals at the level of the cochlea or the cochlear nucleus. Lower level stimuli (50 to 60 dBa) produced a small but significant increase in the ELR amplitude during QS and REM, possibly due to lower interference between the animal's waking activity and the external auditory stimuli. For instance, movement within the recording chamber often produced 50 to 60 dBa levels of noise.

The MLR responses revealed a similar trend during quiet sleep when compared to wake with some important differences. First, lower intensity stimuli produced MLR amplitudes with higher variability and higher mean levels. Since the variability was high, it was difficult to determine if the stimulus intensity and MLR amplitude maintained the same relationship as observed during wake. Surprisingly, the 75 dBa stimulus produced an MLR during QS that was 15 times the size of the 50 dBa stimulus, which may result from the sound evoked post-auricular muscle reflex (PAMR).^47,48 Second, the MLR component during REM sleep remained at a constant low amplitude across stimulus intensity, being severely attenuated at the higher levels. Since REM sleep is characterized by muscle atonia, this muscular reflex and at least some contribution to the MLR, may be inhibited during REM. Thus, while at least some of the MLR has been shown to originate from true neurogenic responses, at higher level stimulus intensities there may be intrusion of the PAMR that masks the neurogenic responses.⁴⁷

From our past experiments, the AEP amplitude characteristics of single intensity, low level stimuli corresponded well to sleep state, with higher LLR corresponding to sleep.^5,42,44 Our present result would suggest that with higher stimulus intensities, this relationship would not be observed, or have the opposite relationship. While other AEP components also differed across state, the LLR component may be most useful for state determination since the differences are more robust. The ELR and MLR components are typically smaller in amplitude, and often with a much lower signal-to-noise ratio. Additionally, over long-term recordings, electrode positions can shift, and AEP amplitude can change based on the relative amount of activated tissue. The current study shows that a paired stimulus with both a high and a low intensity may provide the same information, since the LLR amplitude differences are large during wake and REM, but do not differ in amplitude during sleep. Since individual differences in sensory detection may exist across animals, the LLR relationship provides an additional level of internal control for changes in the ability of the electrode to record the AEP, and could be used to assist in state determination. Additionally, such procedures could be used for different cortical regions (e.g., auditory, somatosensory, and visual), enabling evoked response profiles that may reflect the state of different regions.

In conclusion, auditory stimuli are relatively easy to produce and replicate, potentially making them a reliable tool for investigating state related modulation of brain activity. Since low intensities do not wake the animal,⁴⁴ they can serve as a noninvasive and non-disturbing probe during both sleep and wake. Additionally, auditory AEPs consist of several components reflecting activation of different neural generators in the auditory pathway. The early auditory response may reflect serial activation of different neural generators in the brainstem.¹² Even though the serial activation hypothesis may be an oversimplification of the actual processes involved, it can still provide insight into how behavioral state can influence brain stem structures. As higher order structures are activated, it becomes increasingly difficult to identify specific neural generators involved in the response. However, the constant LLR during QS indicates that this state prevents stimulus intensity modulation of higher level structures. In our previous studies,⁴² we suggested that the AEP amplitude may not be not be helpful in determining whole animal sleep due to the presentation for both high and low amplitude responses during any state. For example, during sleep, the fluctuation between depolarized (wake-like) and hyperpolarized (sleep-like) membrane potential can happen in a delta rhythm with an approximately equal ratio between high and low amplitude responses. This observation leads to the possibility that cortical tissue in particular regions can be in a wake-like state 50% of the time during QS. Such a possibility may lead to a new way to investigate sleep. By using paired stimuli at different intensities with different sensory modalities, an internal control is available to test the input/output relationship for specific tissues, providing a useful metric for whole animal sleep and opens the possibility for probing more localized sleep processes within the cortex.

DISCLOSURE STATEMENT

This was not an industry supported study. The authors have indicated no financial conflicts of interest.

ABBREVIATIONS

AEP

Auditory response potential

dBa

Decibels A-weighted sound pressure

DS

Deep quiet sleep

EEG

Electroencephalogram

EKG

Electrocardiogram

ELR

Early latency response

EMG

Electromyogram

EPSP

Excitatory post synaptic potential

FFT

Fast Fourier transform

ISI

Inter-stimulus interval

LS

Light quiet sleep

LLR

Late latency response

MGB

Medial geniculate body of the thalamus

MLR

Middle latency response

QS

Quiet sleep

REM

Rapid eye movement sleep