Auditory fear conditioning modifies steady-state evoked potentials in the rat inferior colliculus

Local field potential oscillations in the inferior colliculus follow the amplitude envelope of an amplitude-modulated tone, originating a neural response called the steady-state evoked potential. We show that auditory fear conditioning of an amplitude-modulated tone modifies two parameters of the steady-state evoked potentials in the inferior colliculus: first the phase to which the evoked oscillation couples to the amplitude-modulated tone shifts; subsequently, the evoked oscillation power increases along with its coherence with the amplitude-modulated tone.

Abstract

The rat inferior colliculus (IC) is a major midbrain relay for ascending inputs from the auditory brain stem and has been suggested to play a key role in the processing of aversive sounds. Previous studies have demonstrated that auditory fear conditioning (AFC) potentiates transient responses to brief tones in the IC, but it remains unexplored whether AFC modifies responses to sustained periodic acoustic stimulation—a type of response called the steady-state evoked potential (SSEP). Here we used an amplitude-modulated tone—a 10-kHz tone with a sinusoidal amplitude modulation of 53.7 Hz—as the conditioning stimulus (CS) in an AFC protocol (5 CSs per day in 3 consecutive days) while recording local field potentials (LFPs) from the IC. In the preconditioning session (day 1), the CS elicited prominent 53.7-Hz SSEPs. In the training session (day 2), foot shocks occurred at the end of each CS (paired group) or randomized in the inter-CS interval (unpaired group). In the test session (day 3), SSEPs markedly differed from preconditioning in the paired group: in the first two trials the phase to which the SSEP coupled to the CS amplitude envelope shifted ~90°; in the last two trials the SSEP power and the coherence of SSEP with the CS amplitude envelope increased. LFP power decreased in frequency bands other than 53.7 Hz. In the unpaired group, SSEPs did not change in the test compared with preconditioning. Our results show that AFC causes dissociated changes in the phase and power of SSEP in the IC.

NEW & NOTEWORTHY Local field potential oscillations in the inferior colliculus follow the amplitude envelope of an amplitude-modulated tone, originating a neural response called the steady-state evoked potential. We show that auditory fear conditioning of an amplitude-modulated tone modifies two parameters of the steady-state evoked potentials in the inferior colliculus: first the phase to which the evoked oscillation couples to the amplitude-modulated tone shifts; subsequently, the evoked oscillation power increases along with its coherence with the amplitude-modulated tone.

the relevance of sensory cues constantly changes in natural environments, and the adaptive behavior of animals depends on their ability to dynamically modulate the encoding of relevant stimuli. During auditory perception, the coordinated activity of the neuronal assemblies encoding acoustic stimuli results in oscillatory patterns of local field potential (LFP) (Buzsáki et al. 2012; Lakatos et al. 2005). Over the last several years, it has been reported that auditory learning modifies the encoding of behaviorally relevant sounds and influences LFP oscillations in different auditory brain areas (Weinberger 2004). In auditory fear conditioning (AFC), pairing an auditory conditioned stimulus (CS) with a foot shock modifies the neural oscillations in the brain areas involved in CS processing. For example, AFC enhances gamma oscillations in the auditory cortex (Headley and Weinberger 2011) and increases entorhinal cortex-amygdala coherence in the gamma frequency range (Bauer et al. 2007). In the rat auditory midbrain, the inferior colliculus (IC) has been suggested to be an important region for the processing of aversive sounds (LeDoux et al. 1984; Nobre 2013; Nobre et al. 2010). The IC is a major relay for the auditory information ascending to the cortex and amygdala (Ledoux et al. 1987, 1990) and is also a target of feedback projections from these regions (Bajo et al. 2010; Marsh et al. 2002; Winer et al. 2002). However, whether AFC modifies LFP oscillations associated with CS processing in the IC is unknown.

Previous studies explored the effects of AFC on responses evoked by brief CS (1–50 ms) in the IC—a type of response called the transient evoked potential. Mark and Hall (1967) reported that the amplitude of transient LFP deflections evoked by clicks increased in the IC of rats after AFC. In head-fixed bats, Gao and Suga (1998) found that AFC changed the tuning curves of IC toward the fundamental frequency of the CS tone bursts. Of note, transient evoked potentials reflect the sequential activation of successive regions of the auditory system, so the response to one stimulus is finished before the next stimulus is presented (Biacabe et al. 2001; Shaw 1992). In contrast, long-lasting periodic sounds simultaneously evoke a stable oscillatory response at different levels of the auditory system, which phase-locks to the periodic stimulus—a type of response called the steady-state evoked potential (SSEP) (Picton et al. 2003; Stapells et al. 1984). Namely, continuous pure tones with a sinusoidal amplitude modulation have been widely used to induce SSEPs in the auditory system. SSEPs to amplitude-modulated tones are neural oscillations that have the exact same frequency as and phase-lock to the sine wave modulating the amplitude envelope of the pure tone (Herdman et al. 2002; Kuwada et al. 1986, 2002; Rees et al. 1986).

SSEPs may complement the previous findings on transient evoked potentials (Zhang et al. 2013) and provide new insights into the processing of aversive sounds by the neural circuitry of AFC. It has been suggested that the power of the SSEP reflects the excitation of the neurons by afferent synaptic inputs (Pastor et al. 2002) and the coherence of the SSEP to the stimulus amplitude envelope reflects the temporal jittering of such inputs (Kalitzin et al. 2002). The phase to which the SSEP couples to the stimulus amplitude envelope reflects the stimulus-response time lag (Kuwada et al. 2002). Interestingly, human EEG studies demonstrated that auditory learning influences the phase and power of the SSEP (Bosnyak et al. 2004; Gander et al. 2010; Roberts et al. 2012), thus suggesting that associative learning on AFC may also modify the same parameters of the SSEP. However, it remains unexplored whether AFC modifies SSEPs locally generated in the brain regions processing aversive CSs, like the IC.

Here we recorded LFPs from the rat IC during an AFC protocol consisting of five CS trials per day during three consecutive days. The CS was a 30-s 10-kHz tone with a sinusoidal amplitude modulation of 53.7 Hz. In the preconditioning session (day 1), the CS elicited SSEPs that had a peak frequency exactly at 53.7 Hz. After the CS was paired with foot shocks (training session, day 2), striking differences appeared in SSEPs in the test session (day 3) compared with the preconditioning session. The first two test trials showed a ~90° shift in the phase to which the SSEP coupled to the CS amplitude envelope; the last two trials showed an increase in the SSEP power and in the SSEP coherence with the CS amplitude envelope. LFP power increased specifically at 53.7 Hz but decreased in the other frequency bands. In a control group, the foot shocks in the training session occurred randomly in the inter-CS intervals, and the SSEP did not show significant changes in the test. Our results confirm that AFC-related learning modifies SSEP to aversive CS in the IC.

MATERIALS AND METHODS

Ethical approval.

All experiments were conducted in accordance with protocols reviewed and approved by the Institutional Animal Care and Use Committee at the Universidade Federal de Minas Gerais (CEUA-UFMG; protocol no. 133/2010), in accordance with Conselho Nacional de Controle de Experimentação Animal (CONCEA) guidelines defined by Arouca Act 11.794 under Brazilian federal law. CEUA directives are in compliance with National Institutes of Health guidelines for the care and use of animals in research.

Surgery and histology.

Data were obtained from 10 male Wistar rats between 270 and 310 g kept on a 12:12-h light-dark cycle with food and water available ad libitum.

The rats were anesthetized intraperitoneally with a ketamine-xylazine mixture (15 mg/kg and 80 mg/kg, respectively). After pain reflexes were absent, the rats were positioned in a stereotaxic frame (Stoelting, Wood Dale, IL). Supplemental doses of ketamine (20 mg/kg) were administered as needed. After asepsis with povidine-iodine solution (7.5%, topical) and local anesthesia with lidocaine clorohydrate-epinephrine [1% (wt/vol), 7 mg/kg], an incision was made in the scalp to expose the skull. The recording electrode (tungsten microelectrodes, Stoelting; 50-μm diameter) was positioned at the mediolateral coordinates of the of the right IC (−9.0 mm posterior and −1.8 mm lateral to bregma) (Paxinos and Watson 2007). The electrode was slowly lowered into the brain until an evoked response to a tone burst (10 kHz, 50-ms duration) was evident in the raw signal (~3.3 mm ventral to the brain surface). In agreement with the tonotopic organization of the IC—the dorsal layers respond better to low frequencies and the ventral layers to high frequencies (Clopton and Winfield 1973; Huang and Fex 1986)—we histologically verified that the electrode was placed in the ventral part of the IC (see Fig. 2). At the final dorso-ventral coordinate, the electrode was fixed to the skull with zinc cement. A stainless steel screw on the left nasal bone served as the reference electrode for the recordings. Both reference and recording electrodes were connected to the amplifier by means of a RJ-11 connector, which was fixed to the skull with dental acrylic. After the surgery, the rats received prophylactic treatment with pentantibiotics (Zoetis Fort Dodge; 19 mg/kg) and flunixin (Banamine, 2.5 mg/kg) to prevent infection and discomfort and were given 5 days for recovery before recordings were initiated.

Fig. 2.

CS entrains a steady-state evoked potential (SSEP) in the inferior colliculus (IC). A, top: sample histology showing the location of the recording electrode in the IC. Bottom: location of electrode tips across the 10 rats used in this study. B: representative IC LFP response to the amplitude-modulated tone used as CS (light gray, 10-kHz pure tone with a sinusoidal amplitude modulation of 53.7 Hz). Note that CS onset (black arrow) entrains an oscillatory response in the IC LFP that is synchronized to the CS amplitude envelope (CS envelope, purple). Dashed box highlights the time window used in the spectral analysis shown at bottom. Bottom left: IC LFP power spectrum shows a peak at the exact same frequency as the CS envelope (53.7 Hz, red dashed line). Bottom right: phase difference between SSEP and the CS-envelope at 53.7 Hz (SSEP-CS Δ phase). In this example, the LFP and the CS envelope were band-pass filtered between 30 and 70 Hz. C: time course of a representative trial (30-s CS presentation) in the preconditioning session. Top: spectrogram showing SSEP at 53.7 Hz. a.u., Arbitrary units. Middle: SSEP power (red) and SSEP-CS Δ phase (black) at 53.7 Hz. Note stable power increase and stable Δ phase coupling during the trial. Arrows indicate CS onset and offset. Bottom: Δ phase vectors computed for 204.8-ms nonoverlapping windows within the trial (straight lines, left), which were used to estimate the mean phase difference of this trial (right).

After the last recording session, the rats were anesthetized intraperitoneally with urethane [14% (wt/vol), 10 ml/kg] and an electrolytic current (0.5 mA for 2 s) was applied through the recording electrode to generate a small lesion at its tip. The rats were then perfused [intracardiac saline 0.9% (wt/vol) followed by formalin 10% (wt/vol)], and brains were removed and stored in formalin 10%. Parasagittal brain sections (50 µm) were made by cryostat, and the electrode tip was verified by photomicrographs of the histological slices (neutral red staining, × 0.8 magnification) (Fig. 2A).

Auditory stimulus and auditory fear conditioning.

The CS used in our fear conditioning protocol was a 10-kHz pure tone that had its amplitude modulated (100% modulation depth) by a slower sine wave of 53.7 Hz—referred to as CS amplitude envelope (Fig. 2B). The waveforms were generated digitally with a 44.1-kHz sampling rate and 16-bit resolution in Adobe Audition 3.0 and saved to a .WAV file consisting of two channels: a first channel containing the amplitude-modulated tone, which was played through a loudspeaker (tweeter Selenium T-40) at the top of conditioning box as the CS, and a second channel containing only the 53.7-Hz CS amplitude envelope, which was recorded simultaneously along with the neural data. The frequency of the CS amplitude envelope (53.7 Hz) was chosen so that the spectral analysis window (4,096 data samples) contained an integer number of cycles at that frequency, thus avoiding spectral leakage (Felix et al. 2005). Before each behavioral session, the intensity of the CS was adjusted to be at 82 dB SPL (Meeren et al. 2001) and measured at the center of the recording box (Brüel & Kjær type 2238 sound level meter).

AFC took place on three consecutive days in two different contexts, which were differently scented to avoid contextual conditioning effects (Fig. 1A). On the first day (preconditioning), rats were presented with 5 CS trials (30 s each, with 50-s mean interstimulus interval, interval range of 30–120 s) in box 1 (30 × 20 × 25-cm black acrylic box, eucalyptol scented). Before fear conditioning, the rats were randomly assigned to paired (n = 5 rats) or unpaired (n = 5 rats) groups. On the second day (conditioning), rats were presented with 5 CSs (30 s each with 50-s mean interstimulus interval, range 30–120 s) and 5 unconditioned stimuli (USs) in box 2 (23 × 23 × 23-cm gray box, 1% acetic acid scented). The US consisted of a 0.4-mA current applied through metal bars on the floor of box 2 over 2 s. In the paired group US occurred at the last 2 s of the CS, and in the unpaired group US occurred at random moments between CS. On the third day (test), rats were again presented with 5 CSs in box 1, in the exact same conditions as preconditioning.

Fig. 1.

Auditory fear conditioning (AFC) increases freezing to the conditioned stimulus (CS). A: AFC consisted of 5 CS presentations per day in 3 consecutive days. Each CS (light gray) presentation is referred to as a trial and each day as a session. In the preconditioning (left, day 1) and test (right, day 3) sessions, CS trials occurred in box 1. In the conditioning session (center, day 2, box 2), foot shocks (2 s, red) occurred at the end of each CS (paired group, n = 5 rats) or were randomized in the inter-CS interval (unpaired group, n = 5 rats). B: mean freezing (±SE) to CS in preconditioning and test sessions. Note that freezing increases in the test session in the paired group. C: mean freezing (±SE) to CS by trial number. ^#P < 0.05 preconditioning vs. test, *P < 0.05 paired vs. unpaired, 2-way ANOVA followed by Bonferroni.

The behavioral sessions were recorded by a video camera, and the videos were analyzed off-line by an examiner blind to the condition. Freezing behavior was defined as the absence of movement of the rats, except breathing, for a minimum of 3 s and was expressed as the percentage of time freezing during each CS trial (Blanchard and Blanchard 1972; Curzon et al. 2009; Fanselow and Bolles 1979). Periods outside of the CS presentation were not considered for freezing quantification.

Electrophysiological recordings and data analysis.

LFPs were recorded from the ventral part of the central nucleus of IC in preconditioning and test sessions. The signals were filtered between 1 and 2,000 Hz, amplified by 20,000 V/V (Cyberamp 320; Axon Instruments), and then digitalized (NI-DAQ 6023E; National Instruments; 12-bit resolution, ±5 V range, and sampling rate of 20 kHz). CS amplitude envelope was recorded with the same filter parameters and amplified by 1 V/V. Recorded data were analyzed off-line with built-in and custom-written MATLAB codes (MATLAB 7.12 R2011a). In Fig. 2B and Fig. 6B, the representative time series were obtained by band-pass filtering the raw LFPs between 30 and 70 Hz with the eegfilt.m function. In other cases, analyses were carried out on unfiltered LFPs. Time-frequency power and phase decomposition were calculated by means of a built-in spectrogram function (nonoverlapping, 4,096-point Hamming window), which yielded complex frequency (f) domain representations of the signals at each 204.8-ms time (t) step:

Fig. 6.

Time course of SSEP in early and late test trials. A: representative SSEP power (red) and SSEP-CS Δ phase (gray) in the 1st and 4th test trials of paired rat. In 1st trial, Δ phase shifts in the last 10 s but not in the beginning of the trial. Note that different angles of Δ phase are not associated with changes in SSEP power. In the 4th trial, SSEP power increases and no Δ phase shift occurs. Arrows indicate CS onset and offset. B: segments of IC LFP (gray) and CS envelope (purple) obtained from the periods indicated by arrowheads in A. Power spectra (red) and Δ phase (light gray) of these segments are depicted at bottom. LFPs and CS envelope were band-pass filtered between 30 and 70 Hz. C and E: mean SSEP power and mean SSEP-CS Δ phase (±SE), respectively, in early and late test trials in paired group (n = 5). Pink marks indicate periods of 10 s (a, b, c). Arrows indicate CS onset and offset. D and F: mean SSEP power and mean SSEP-CS Δ phase (±SE) of 1st preconditioning trial and of periods a, b, and c (n = 5 rats). ^#P < 0.05 in comparison to preconditioning, 1-way ANOVA followed by Bonferroni.

$$S\left(f,t\right)=A\left(f,t\right)\times e^{\left(f,t\right)}$$

SSEP-CS Δ phase was calculated in each time step as the difference between the imaginary components of LFP and CS amplitude envelope at f = 53.7 Hz. The zero Δ phase for each rat was defined as the mean phase of its preconditioning trials. This normalization procedure is in accordance with previous works (Besle et al. 2011; Gander et al. 2010). Each trial had N unit normalized SSEP-CS Δ phase vectors, and the sum of such vectors provided two variables:

1) mean Δ phase: argument of the sum of the Δ phase vectors:

$$\frac{1}{N}\text{arg}\left[{\displaystyle \sum _{n=1}^N{e}^{(n)}}\right]$$

2) coherence: absolute value of the sum of the Δ phase vectors:

$$\frac{1}{N}\left|{\displaystyle \sum }_{n=1}^N{e}^{\left(n\right)}\right|$$

Coherence is an adimensional real value between 0 and 1 and is a metric of SSEP-CS phase coupling in each trial. Considering a circular distribution of Δ phase vectors during CS presentation, coherence = 0 means uniform Δ phase distribution and coherence = 1 means perfect Δ phase grouping (Kalitzin et al. 2002; Womelsdorf et al. 2007).

The power of the SSEP was also quantified for each trial by multiplying the Fourier coefficients by its complex conjugate. The power of each trial was then divided by the mean power of preconditioning trials for normalization.

Total LFP power was calculated by integrating the power from 1 to 90 Hz (the 46–66 Hz frequency range was excluded from this analysis to avoid quantifying SSEP components). LFP frequency bands were defined as theta (5–10 Hz), alpha (11–16 Hz), beta (17–29 Hz), and gamma (30–46 Hz and 66–90 Hz) and power quantified by integration of the respective frequency band.

For calculation of group means in Fig. 6, the test trials that showed the highest Δ phase shift (1st trial in 4/5 rats but 2nd trial in rat 4; see Fig. 3B) and the trials that showed the lowest Δ phase shift (5th trial in 4/5 rats but 4th trial in rat 5) were selected. To carry out statistical comparisons in such trials, they were divided into three periods of 10 s each (depicted as a, b, and c in Fig. 6C).

Fig. 3.

SSEP-CS Δ phase shifts after AFC in early test trials. A: SSEP-CS Δ phase in representative paired and unpaired rats. Δ phase was stable across the 5 preconditioning trials in paired and unpaired rats; in test, Δ phase in paired rats shifted in the 1st and 2nd trials and returned to preconditioning values in the late trials. Unpaired rats showed no Δ phase shifts in test. B: test trials of occurrence of the highest and the lowest Δ phase shifts in paired rats. Note that the highest Δ phase shifts occurred in early trials while the lowest shifts occurred in late trials. Unpaired group is not depicted as it showed no significant shifts. C: mean SSEP-CS Δ phase (±SE) across trials (n = 5 rats/group). ^#P < 0.05 preconditioning vs. test, *P < 0.05 paired vs. unpaired, 2-way ANOVA followed by Bonferroni.

Statistical analysis.

All data are presented as means ± SE. The normal distribution of the data was confirmed by the Kolmogorov-Smirnov test (P > 0.10). Statistical analyses were performed with one-way or two-way repeated-measures analysis of variance (ANOVA), followed by Bonferroni post hoc test, in accordance with the coefficient of variation of the data set. Statistical significance was assumed when P < 0.05. Statistical analysis was carried out in GraphPad Prism 6.0 and MATLAB 7.12 R2011a (The MathWorks, Natick, MA).

RESULTS

Behavioral response to amplitude-modulated CS.

It has been reported that AFC induces the emergence of defensive behavior to an aversive CS along with changes in LFP oscillations in the auditory system (Bauer et al. 2007; Collins and Paré 2000; Rogan et al. 1997; Weinberger et al. 2013). Here we used an amplitude-modulated tone as the CS in an AFC protocol (5 CSs per day in 3 consecutive days) and simultaneously recorded IC LFP responses to such CS during preconditioning (day 1) and test (day 3) sessions (Fig. 1A; Fig. 2, A and B). In the training session (day 2), we applied foot shocks at the end of each CS (paired group, n = 5) or randomized in the inter-CS interval (unpaired group, n = 5). In the preconditioning session, the CS elicited low levels of freezing in both groups (day 1, Fig. 1, B and C), indicating that it was initially not aversive. Freezing in the paired group significantly increased in the test session (day 3, Fig. 1B, F_1,8 = 20.33, P < 0.001)—note that it intermediately increased in the first trial and achieved levels near 100% in the ensuing trials (Fig. 1C, F_9,72 = 5.75, P < 0.01 and P < 0.0001, respectively). Importantly, the unpaired group maintained low freezing levels in the test. As shown in Fig. 1, B and C, the comparison between the paired and unpaired groups yielded significant differences in the test session (P < 0.05).

SSEP-CS Δ phase shifts in early test trials.

Amplitude-modulated tones entrain neural oscillations in the auditory pathway, which are phase-coupled to the amplitude envelope of the acoustic stimulus (Kuwada et al. 1986; Rees et al. 1986). Accordingly, the CS we used here—a 30-s 10-kHz tone with the amplitude envelope modulated by a 53.7-Hz sine wave—evoked prominent LFP oscillations in IC, which were synchronized to the 53.7-Hz amplitude envelope (Fig. 2B). Hence we keep the terminology of previous studies and refer to this evoked LFP oscillation as the “steady-state evoked potential” (SSEP) (Picton et al. 2003; Stapells et al. 1984). During CS presentation in preconditioning, SSEP was evident in the IC LFP power spectrum as a peak frequency exactly at 53.7 Hz; also, the cycles of SSEP coupled to the CS amplitude envelope at a specific phase angle, referred to as SSEP-CS Δ phase (Fig. 2B). It is noteworthy that time-resolved frequency analysis of LFPs evidenced that SSEP power remained stable during CS presentation; similarly, SSEP-CS Δ phase showed small variation around a mean phase angle (the mean Δ phase of the trial) (Fig. 2C).

SSEP-CS Δ phase is a metric of the mean time lag between the CS amplitude envelope and the IC response (Kuwada et al. 2002) and has been reported to be affected by auditory learning (Besle et al. 2011; Gomez-Ramirez et al. 2011). In accord with the neutral valence of the CS before AFC, we did not observe significant differences in the SSEP-CS Δ phase values between preconditioning trials (Fig. 3, A and C). Interestingly, the paired group showed pronounced Δ phase shifts in the first and second test trials of the CS (Fig. 3, B and C; group mean Δ phase of −69.6° and −76.9°, respectively), which were significantly different from preconditioning (F_9,72 = 2, P = 0.0018; P < 0.05 and P < 0.01, respectively). Of note, in the ensuing test trials Δ phase showed no significant shifts in comparison to preconditioning. Consistently with our behavioral findings, the unpaired group showed no significant Δ phase shifts. When comparing between paired and unpaired groups, we observed significant differences in SSEP-CS Δ phase only in the second test trial (Fig. 3C, P < 0.01). These results indicate that AFC determines phase resetting of SSEP, although this effect is limited to the early test trials.

SSEP power and SSEP-CS phase coherence increase in late test trials.

We next assessed the synchronization of SSEP to the CS amplitude envelope by means of phase coherence (Kalitzin et al. 2002; Womelsdorf et al. 2007), which is a metric of the clustering/dispersion of SSEP-CS Δ phase vectors in each trial (see materials and methods). Polar plots in Fig. 4A show Δ phase vectors of example paired and unpaired rats across CS trials in the preconditioning and test sessions. Interestingly, SSEP-CS Δ phase coherence in the paired group increased only in the late test trials—in marked contrast to the Δ phase shifts, which occurred in early test trials. Group means confirmed that phase coherence in the fourth and fifth test trials (0.93 and 0.94, respectively) were significantly higher compared with preconditioning (0.63) and with the unpaired group (0.57 and 0.50, 4th and 5th trials, respectively) (Fig. 4B, F_6,24 = 8; P = 0.01 vs. preconditioning, P < 0.05 vs. unpaired).

Fig. 4.

SSEP-CS phase coherence increases in late test trials. A: polar plots of SSEP-CS Δ phase and phase coherence (above polar plots) calculated for each trial in representative paired and unpaired rats. Only the 1st preconditioning trial is depicted. Null values indicate uniform distribution of Δ phase vectors in the unitary circle, while 1 indicates perfect clustering of Δ phase vectors to a specific angle. Note in the 4th and 5th trials of the paired rat the high clustering of Δ phase vectors associated with coherence near 1. In the 1st trial of the paired rat the shift of Δ phase vectors is evident, though not associated with changes in coherence. B: mean coherence (±SE) across trials (n = 5 rats/group). ^#P < 0.05 preconditioning vs. test, *P < 0.05 paired vs. unpaired, 2-way ANOVA followed by Bonferroni.

Supporting that AFC enhances the entrainment of IC to the aversive CS, the response gain increased during the test session. In the paired group, the SSEP power increased in the fourth and fifth test trials (Fig. 5, A and B), closely resembling what we have observed for the phase coherence. In the fourth and fifth test trials, the mean SSEP power in the paired group was significantly higher than preconditioning (Fig. 5B, F_9,72 = 3.284, P = 0.0021 and P < 0.05, respectively) and the SSEP power in the unpaired group (P < 0.001).

Fig. 5.

SSEP power increases in late test, but total LFP power decreases in early and late test. A: time-frequency power analysis of IC LFPs in representative paired and unpaired rats across AFC trials. CS depicted at top in gray. Only the 1st preconditioning trial is depicted. Note that SSEP power in the paired rat markedly increased in 5th test trial. Unpaired group did not show significant SSEP power changes in test. B: mean SSEP power (±SE) across trials (n = 5 rats/group). ^#P < 0.05 preconditioning vs. test, *P < 0.05 paired vs. unpaired, 2-way ANOVA followed by Bonferroni. C: total LFP power (1–46 Hz and 66–90 Hz) before CS onset (no CS, square) and during CS presentations (CS, circle). In the test session, CS presentation reduced total LFP power in the paired group, but no significant effect occurred in the unpaired group. Note that decreases in total LFP power in the paired group are similar in early and late test trials. D: LFP power discriminated by frequency bands. All frequency bands of LFP showed significant power decrease in the paired group. Groups represented as means ± SE, n = 5 rats/group. ^#P < 0.05 in comparison to no CS Pre; *P < 0.05 in comparison to unpaired group; 2-way ANOVA followed by Bonferroni.

To confirm that AFC increased LFP power specifically at the SSEP frequency (53.7 Hz), we carried out an additional analysis to quantify power in the other frequency bands of the IC LFP spectrum (1–46 Hz and 66–90 Hz, to exclude SSEP frequency). In the paired group, the presentation of the aversive CS in the test trials significantly decreased total LFP power in comparison to the preconditioning trials. Differently from the AFC effects in SSEP power—which occurred only in the fifth trial—the decrease in total LFP power occurred in all the test trials (Fig. 5C, F_3,12 = 3, P < 0.001) and appeared in all the frequency bands of the LFP spectrum (Fig. 5D; theta: F_6,24 = 9, alpha: F_6,24 = 7, beta: F_6,24 = 6, gamma: F_6,24 = 11; P < 0.001). The unpaired group showed no significant changes in total LFP power. Therefore, the enhanced entrainment of the IC LFPs to the aversive CS—increased SSEP power and increased SSEP-CS Δ phase coherence—is an effect of AFC observed in the late test trials, being dissociated from the Δ phase shifts observed in early test and from the decreases in total LFP power.

Time course of SSEP power and Δ phase in early and late test trials.

The results reported so far point out to differential effects of AFC in SSEP phase and power in early and late test trials. We next proceeded to analyze the time course of individual test trials to assess how the shifts in Δ phase and the increases in power emerged within trials. Figure 6A shows time-resolved LFP power and phase analysis at 53.7 Hz in representative early and late test trials. In the early test trial, it is noteworthy that the Δ phase coupled to different angles when we compare the first 10 s to the last 10 s. While the first 10 s exhibited a coupling pattern resembling preconditioning trials (Δ phase ~0°), the last 10 s showed pronounced Δ phase shifts to ~90°. The different modes of phase coupling are evident at the inspection of the LFPs and CS amplitude envelope time series at different moments of the early trial (Fig. 6B). Group means confirmed that the Δ phase in the last 20 s of the early test trial significantly differed from the Δ phase in preconditioning. The Δ phase in first 10 s of the early trials, however, did not differ from preconditioning (Fig. 6, E and F, F_6,24 = 8, P < 0.001; P < 0.01 in b, P < 0.01 in c). Interestingly, the different modes of Δ phase coupling in the early test trials—to ~0° or to ~90°—did not influence SSEP power (Fig. 6, C and D). This finding rules out the possibility that SSEP power increase was already established in the early tests but was masked by the phase shifting. Instead, we found that SSEP power only increased in the late test trials (F_6,24 = 7, P = 0.0003; P < 0.01 in b, P < 0.05 in c), coinciding with the lowest Δ phase shifts observed in the test.

DISCUSSION

In this study we investigated whether AFC modifies SSEP in IC LFPs. By comparing the phase and power of SSEP on a trial-by-trial basis, here we have shown distinct effects of AFC on test session in comparison to preconditioning. In early test (trials 1 and 2), we found shifts in the SSEP-CS Δ phase; in late test (trials 4 and 5), we found increased entrainment of SSEP to the CS—measured by increases in SSEP-CS coherence and in SSEP power. Furthermore, our results show that the early test trials exhibit two modes of SSEP-CS phase-coupling: no Δ phase shifts in the first several seconds, followed by prominent Δ phase shifts in the last seconds. These different modes of phase coupling were not associated with changes in power, which we only observed in the late test.

Our findings confirm previous studies that showed that AFC modifies the neural processing of aversive CSs in the IC (Gao and Suga 1998; Ji and Suga 2009; Mark and Hall 1967). While these studies used CSs ranging from 1 to 50 ms to evoke transient responses, we used a 30-s amplitude-modulated tone to entrain SSEPs in IC LFPs. In contrast to transient evoked potentials, the detection of SSEPs is not based on an average of trials. As the SSEPs had a precise frequency (53.7 Hz), its detection—and its differentiation from the endogenous oscillatory activity—was straightforward in the LFP power spectrum (Fig. 2, B and C). Hence, we could assess IC responses on a trial-by-trial basis (Figs. 3–5) as well as the time course of the responses (Fig. 6)—which would not be possible using transient responses. Another drawback of transient stimulation is the abrupt changes in the auditory scene, which automatically activate arousal circuits (Aston-Jones and Bloom 1981; Folk and Remington 2015; Remington et al. 1992). Long-lasting sounds used to elicit SSEP, in contrast, create a stable acoustic environment and are less susceptible to the effects of stimulus salience (Yantis and Jonides 1996). In fact, Ji and Suga (2009) reported AFC-related changes to a transient CS in their unpaired control group; also, Mark and Hall (1967) reported nonspecific effects in their experiments, for example, increased responses to transient visual stimuli after AFC. Importantly, our unpaired control group did not show changes in freezing behavior or in SSEP in the test session of AFC. Thus the behavioral and electrophysiological changes in the paired group resulted from learning the association between the CS and the foot shock.

Long-lasting amplitude-modulated tones simultaneously entrain the brain stem and cortical areas of the auditory system, and these auditory regions couple to specific phases of the amplitude-modulated tones (Kuwada et al. 2002). Galambos and Makeig (1992) suggested that the SSEP-CS Δ phase in successive auditory regions corresponds to a fixed latency of activation by feedforward inputs. Our finding of phase shifts after learning opposes the hypothesis of Galambos and Makeig and corroborates previous SSEP studies (Besle et al. 2011; Gander et al. 2010) that also reported shifts of SSEP-CS Δ phase. One possibility is that the resultant SSEP-CS Δ phase reflects the interaction of feedback and feedforward circuitry in the IC during sound processing, although further experiments are needed to evaluate this hypothesis. Indeed, recent findings on cortical circuits suggest that the phase relations of oscillations in distant brain regions mediate their interaction and the routing of information through sensory circuits (Besserve et al. 2015; Lakatos et al. 2008; Womelsdorf et al. 2007). We extended these findings by showing that phase shifts also occur in the auditory midbrain.

We found that AFC-induced changes in SSEP Δ phase occurred in the early test trials, followed by changes in SSEP power in the late test trials. Previous studies showed that increased responsiveness to the CS appears sequentially in the different levels of the auditory pathway across test trials. Quirk et al. (1997) reported that the firing rates in the lateral amygdala increased in early test trials, while in the auditory cortex it only increased in late trials. Disterhoft and Stuart (1976) showed that, across test trials of a reward-predicting CS, firing rates increased first in the cortex, followed by the thalamus, and subsequently by the brain stem. Our results are in accord with these findings, as we have shown that SSEP power in the IC increases in late, but not in the early, test trials. Thus the phase shifts we observed in the early test trials may coincide with an increase in the activity of prosencephalic brain regions processing the CS, which could influence IC activity by means of feedback projections (Bajo and King 2013). A recent study showed that an animal model of audiogenic seizures, which has a pathological hyperexcitability of the acoustic-limbic pathway, also exhibited marked changes in phase and power of IC SSEPs compared with control rats (Pinto et al. 2017). Some theories indeed predict that feedback projections may influence the activity in early stages of sensory systems to stabilize the encoding of relevant stimuli (Grossberg 1980, 1999). Interestingly, the integrity of the cortex and amygdala is necessary for the development of plasticity in subcortical auditory regions (Bajo et al. 2010; Maren et al. 2001).

Additionally, we found that phase coherence increased in the late test trials. The spread/clustering of Δ phase values around their mean—as measured by SSEP-CS coherence—has been suggested to reflect the degree of jittering between two oscillators (in our case, between the amplitude-modulated tone and the SSEP) (Kalitzin et al. 2002). It is important to remark that even though coherence and power increased together in late trials, coherence is a metric that depends only on the stability of Δ phase during the time course of the trial—and is independent of the power and of the mean Δ phase (see materials and methods). Thus AFC stabilizes the encoding of the CS in the IC by enhancing the excitability (power increase) and reducing the jittering (coherence increase) of the neural responses to the aversive CS.

Recent studies showed that AFC enhances gamma oscillations in the auditory cortex in response to an auditory CS (Headley and Weinberger 2011; Weinberger et al. 2013). In these studies, the CS was a nonmodulated pure tone and gamma oscillations emerged in a wide frequency band (40–120 Hz) as an intrinsic property of auditory cortex circuits. The SSEP we report here is also an oscillation in the gamma frequency range, but it is generated by a different mechanism—the entrainment of the IC to the amplitude envelope of the CS. Indeed, we found that the power of the intrinsic gamma band (30–46 Hz and 66–90 Hz) decreased in all the test trials, while the power at 53.7 Hz increased specifically in the fifth and fourth test trials (Fig. 5). This latter result reinforces that different network mechanisms generate intrinsic gamma oscillations and SSEP oscillations; it also indicates that changes in intrinsic gamma cannot account for changes in SSEP Δ phase and power.

In conclusion, our results confirm that associative learning influences the phase and power of SSEPs locally recorded from the rat IC. In agreement with previous studies (Bosnyak et al. 2004; Gander et al. 2010; Roberts et al. 2012), our findings suggest that phase and power of SSEP reflect distinct aspects of neural processing of sounds, which are subject to specific mechanisms of plasticity emerging at different moments of the test. These findings indicate that SSEPs may provide important insights into the neural processing of aversive CS by neural circuitry of the AFC, which were previously unexplored by transient evoked potentials.

GRANTS

This work was supported by Fundação de Amparo à Pesquisa do Estado de Minas Gerais (CBB-APQ-02290-13 and TEC-APQ-01084-13), Coordenação de Aperfeiçoamento de Pessoal de Nível Superior (MINCYT 0951/2013), and Conselho Nacional de Desenvolvimento Científico e Tecnológico (476681/2012-0, 470532/2012-2, and 306767/2013-9).

DISCLOSURES

No conflicts of interest, financial or otherwise, are declared by the authors.

AUTHOR CONTRIBUTIONS

A.L.V.L., F.A.G.M., and M.F.D.M. conceived and designed research; A.L.V.L. and F.A.G.M. performed experiments; A.L.V.L. analyzed data; A.L.V.L., F.A.G.M., and M.F.D.M. interpreted results of experiments; A.L.V.L. and F.A.G.M. prepared figures; A.L.V.L., F.A.G.M., and M.F.D.M. drafted manuscript; A.L.V.L., F.A.G.M., and M.F.D.M. edited and revised manuscript; A.L.V.L., F.A.G.M., and M.F.D.M. approved final version of manuscript.