RGFP966

Inhibition of histone deacetylase 3 via RGFP966 facilitates cortical plasticity underlying unusually accurate auditory associative cue memory for excitatory and inhibitory cue-reward associations

Authors: Andrea Shang, Sooraz Bylipudi, Kasia M. Bieszczad PII: S0166-4328(18)30237-7
DOI: https://doi.org/10.1016/j.bbr.2018.05.036
Reference: BBR 11457

To appear in: Behavioural Brain Research

Received date: 14-3-2018
Revised date: 29-5-2018
Accepted date: 30-5-2018

Please cite this article as: Shang A, Bylipudi S, Bieszczad KM, Inhibition of histone deacetylase 3 via RGFP966 facilitates cortical plasticity underlying unusually accurate auditory associative cue memory for excitatory and inhibitory cue-reward associations, Behavioural Brain Research (2018), https://doi.org/10.1016/j.bbr.2018.05.036

This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

Inhibition of histone deacetylase 3 via RGFP966 facilitates cortical plasticity underlying unusually accurate auditory associative cue memory for excitatory and inhibitory cue-reward associations.

Abstract

Epigenetic mechanisms are key for regulating long-term memory (LTM) and are known to exert control on memory formation in multiple systems of the adult brain, including the sensory cortex. One epigenetic mechanism is chromatin modification by histone acetylation. Blocking the action of histone de-acetylases (HDACs) that normally negatively regulate LTM by repressing transcription, has been shown to enable memory formation. Indeed, HDAC-inhibition appears to facilitate memory by altering the dynamics of gene expression events important for memory consolidation. However less understood are the ways in which molecular-level consolidation processes alter subsequent memory to enhance storage or facilitate retrieval. Here we used a sensory perspective to investigate whether the characteristics of memory formed with HDAC inhibitors are different from naturally-formed memory. One possibility is that HDAC inhibition enables memory to form with greater sensory detail than normal. Because the auditory system undergoes learning-induced remodeling that provides substrates for sound-specific LTM, we aimed to identify behavioral effects of HDAC inhibition on memory for specific sound features using a standard model of auditory associative cue-reward learning, memory, and cortical plasticity. We found that three systemic post-training treatments of an HDAC3-inhibitor

(RGPF966, Abcam Inc.) in rats in the early phase of training facilitated auditory discriminative learning, changed auditory cortical tuning, and increased the specificity for acoustic frequency formed in memory of both excitatory (S+) and inhibitory (S-) associations for at least 2 weeks. The findings support that epigenetic mechanisms act on neural and behavioral sensory acuity to increase the precision of associative cue memory, which can be revealed by studying the sensory characteristics of long-term associative memory formation with HDAC inhibitors.

Highlights

1.Inhibition of HDAC3 facilitates acquisition of associative sound discriminations.

2.Inhibition of HDAC3 enables faster behavioral responses to appropriate cues.

3.HDAC3 negatively regulates the precision of sound-specific long-term memory.

4.HDAC3 negatively regulates long-lasting sound-specific cortical tuning changes.

5.Post-training RGFP966 treatment early in training has lasting behavioral effects.

Keywords

epigenetics, HDACs, memory, discrimination, associative learning, auditory cortex

1.Introduction

Learning and memory research has long been focused on investigations of the mechanisms controlling the strength of memory formation over time and interference. In this domain, research findings support that gene expression is necessary for the formation of long-term memory (LTM) (Alberini, 2009 [1]), defined here as that which lasts over time usually beyond ~24 hours. Thus, molecular mechanisms that control experience- dependent gene expression during memory consolidation have been investigated as molecular-level requisites for successful experience-dependent plasticity underlying the formation of new LTM. Recent research in the neurobiology of learning and memory has highlighted the importance of epigenetic control over experience-dependent gene expression in the adult brain.
Epigenetic chromatin remodeling occurs by various mechanisms including, but not limited to, DNA methylation, histone acetylation and nucleosome remodeling (Sweatt, 2013 [2]). Here, the focus is on histone acetylation, which regulates gene expression via enzymes called histone acetyltransferases (HATs) and histone deacetylases (HDACs) that together epigenetically regulate novel bouts of gene expression required for LTM (reviewed in Barrett & Wood, 2008 [3]). HATs tend to relax chromatin structure, usually enabling experience-dependent gene expression, whereas HDAC enzymes oppose acetylation to restrict chromatin and usually repress activity-dependent transcription. Therefore, promoting HAT function while blocking the action of HDACs alters chromatin to permit an “open” state of euchromatin that is permissive to transcription initiated by a learning experience and required for subsequent formation of long-term memories.

In this respect, HDACs have been shown to be powerful negative regulators of LTM formation. In particular, the inhibition of HDACs is known to facilitate memory processes for learning events that would otherwise not have been later remembered. Strikingly, LTM enabled by such HDAC inhibition produces memories that persist beyond the time point at which normal memory would have failed (Stefanko et al., 2009 [4]). This suggests that memories induced with HDAC inhibition last longer than normal LTM. These induced memories will be subsequently referred to as unusually persistent memory, or UPM. The persistence of UPM suggests that their underlying neural substrates are somehow different to those of long-term memories formed under normal learning and consolidation conditions. Initial interpretations of the surprising discovery of UPMs was that the memory formed was stronger than normal, with strength defined as an increased durability to the passage of time, likely by increased synaptic potentiation in memory circuits (e.g., as increased LTP in the hippocampus; Shu et al., 2018 [5]). An alternative and complementary interpretation is that UPMs are more robust to time or interference due to the incorporation of more perceptual details encoded into memory, which would involve substrates from a larger neural network. Since the remembered sensory details of an experience provide more potential environmental features to cue and trigger retrieval, UPMs would gain their robustness over typical LTM because they contained more of the detailed cues that could later be used to activate recall and motivate behavior. This leads us to the hypothesis that memories formed with HDAC inhibition have transformed more sensory features from a learning experience into what becomes the contents of the memory for that experience. This potential explanation of HDAC function in memory is formally proposed as the “informational capture hypothesis” (Bieszczad et al., 2015 [6]; Phan &

Bieszczad, 2016 [7]). The idea is based on the “molecular brake pad hypothesis” set forth

by McQuown & Wood, (2011) [8], which states that HDACs usually act as molecular

“brakes” on gene expression required for memory formation. Taken together, these hypotheses link molecular- with systems-level neural mechanisms of memory formation. The proposal is that HDAC inhibition releases molecular brakes on systems-level plasticity so that recently active systems-level representations of an experience become transformed to create a legacy of that experience, rather than remaining stable. If such a process for cue- specific plasticity is initiated in a sensory system, then HDAC inhibition would be predicted to facilitate discriminative cue learning. Thus, the learning curve of an incremental auditory associative discrimination would be expected to steepen with HDAC inhibition because of better maintenance in the auditory system of the neural impressions of the particular acoustic features of the discriminative cues being learned from one day to the next. A recent paper by Malvaez et al., (2018 [9]) has suggested subtle but significant effects on incremental instrumental learning itself using HDAC inhibitors (see their Figure 1B and C; but see also their Figure 3D, 3E). Therefore, the focus here is to parse out and identify a sensory role for HDAC-mediated changes in the incremental acquisition and retention of associative cue memory and behavior.
Indeed, not all experiences that activate sensory population responses induce stimulus-specific plasticity, and not all learning events transform into memories, as is the case for non-salient or behaviorally insignificant experiences (Recanzone et al., 1993 [10]; Rutkowski & Weinberger, 2005 [11]; Berlau & Weinberger, 2008 [12]; Bieszczad &
Weinberger, 2010 [13]; McGaugh, 2013 [14]). Thus, HDAC inhibition may permit LTM and enable UPM by increasing the amount and detail of sensory features that are encoded

from a learning experience into a memory of that experience regardless of motivational affect. This may be achieved by a kind of sensory capture enabling plasticity in more of the transiently activated sensory network during perception, without regard to normal associative learning rules of valence, motivation, or salience that typically drive plasticity events (Kwapis & Wood, 2014 [15]; White & Wood, 2014 [16]).
Sensory system models are particularly well-suited as an approach to investigate whether the specific sensory “content” of associative cues become encoded because of the relative ease of determining their neural representations after the formation of associative memory (Scheich et al., 2007 [17]; de Villers-Sidani & Merzenich, 2011 [18]; Grosso et al., 2015 [19]). Sensory cortical plasticity is observable in the sensory receptive fields of cells within cortical systems that are organized in an experimentally accessible topographic manner. This spatial organization—like for acoustic frequency in the tonotopy of primary auditory cortex (A1)—allows access to the neural representation of the particular stimulus that can also be tested at a behavioral level. For example, the associative sound cue can be linked to its behavioral outcome (e.g., go or no-go) by the shape of its receptive field and cortical representation in the auditory cortex. Further, plasticity in the tonotopic representation of A1 is known to induce lasting changes in neural activity that remodels the cortical representation of specific sound-frequencies with learned behavioral relevance (e.g., Rutkowski & Weinberger, 2005 [11]; reviewed in Weinberger, 2015 [20]). As such, frequency-specific auditory cortical remodeling is related to the formation of strong memory for specific sound cues (e.g. Bieszczad & Weinberger, 2010 [13]; 2012 [21]). Accumulating evidence suggests that representational plasticity across A1 is a likely initial neural substrate of the specific acoustic details available for later storage or retrieval in

long-lasting auditory associative memories and serves as a model for sensory system substrates of memory in any modality (Ghose, 2004 [22]; Ohl & Scheich, 2005 [23]; Weinberger, 2015 [20]; McGann, 2015 [27]).
By extension from known neurobiological correlates in auditory cortical tonotopic plasticity, behavioral studies of memory in an auditory model are uniquely positioned to address the potential role of epigenetic mechanisms to regulate the specific acoustic contents of memory. Furthermore, prior work in the auditory cortex has already implicated HDAC3 as a critical negative regulator of auditory cortical plasticity and sound-specific memory formation for sound frequency (Bieszczad et al., 2015 [6]). Rats treated with RGFP966 (the class I HDAC-inhibitor with selectivity for HDAC3 used in the present study; introduced in Malvaez et al., 2013 [25]) while learning a simple single sound-reward association formed more acoustically specific memory for the behaviorally-significant sound frequency. Moreover, A1 became unusually “tuned-in” to the specific sound frequency and sound level features that could be associated with reward (compared to vehicle-treated subjects). This supports a strong link between HDACs, sensory-specific cortical plasticity and cue-specific memory. Whether epigenetic mechanisms regulate how much detailed sensory information from learning experiences is encoded to memory to drive discriminative behavior remains an open question. Recent findings using the same RGFP966-mediated inhibition of HDAC3 in an avian model confirmed the importance of HDAC3 for inducing feature-specific auditory neuroplasticity after experiences with complex auditory vocalization stimuli with exceptionally detailed acoustic features (Phan et al., 2017 [26]). That sound-specific auditory neural plasticity processes in both mammals and birds is enabled by HDAC3-inhibition suggests an evolutionarily conserved role for

HDACs to regulate the sensory details consolidated to memory through neural plasticity mechanisms. However, whether HDACs alter the rapid acquisition of new sensory information is unknown. Many tasks used in this domain of research, as in Phan et al., (2017 [26]), use a single-session and non-incremental task in which learning effects in acquisition are difficult to ascertain. In addition, Bieszczad et al., (2015 [6]) found no change in the learning curve of animals treated with HDAC3-inhibtion while acquiring a single tone association with reward. The purpose of this experiment was to identify whether learning effects would be evident if the task required an active associative discrimination between two tones: one sound frequency associated with reward, and another frequency, without reward.
In order to determine the influence of HDAC3 on the sensory characteristics of memory formed, we designed a behavioral instrumental conditioning and discrimination task that challenged rats to learn and remember two sound-frequency associations for reward outcomes over many days of training: One sound was an excitatory association by contingency with reward to elicit a go response (“conditioned stimulus plus”, S+) and a second, easily distinguishable sound was an inhibitory association with reward, i.e., predicted “no reward” for a no-go response (the S-). The HDAC3-selective pharmacological inhibitor (RGFP966, Abcam Inc.,) was used to block HDAC3 in immediate post-training “consolidation” time periods (McGaugh, 1989 [27]). Unlike Bieszczad et al., (2015 [6]), the HDAC3-inhibitor treatments were limited to the early phase of learning this incremental task (i.e., after only the first 3 training sessions). If HDAC3-inhibition enhances LTM formation by enabling specific sensory details in
memory (supporting the so-called informational capture hypothesis), then behavioral

evidence of UPM would be revealed by facilitated acquisition of the two-tone discrimination task within days of treatment with the HDAC3-inhibitor, RGFP966. Furthermore, rats treated with RGFP966 would be expected to also show more sound- specific memory revealed by behavioral response peaks in a sound generalization memory test for not just one, but both of the two behaviorally relevant acoustic-frequencies (compared to vehicle-treated controls) after acquiring the task. We found behavioral evidence for both of these effects of systemic HDAC3-inhibition by RGFP966, along with evidence of auditory cortical plasticity for specific sound frequency in animals treated with the inhibitor. The results support a novel role for HDAC3 during auditory memory acquisition, including a confirmation of consolidation and cortical plasticity effects, that regulate the neural and behavioral acoustic specificity of newly learned sound-signal associations.

2.Methods and Materials

2.1.Subjects

A total of 24 adult male Sprague-Dawley rats (275-300g on arrival; Charles River Laboratories, Wilmington MA) were used in behavioral and electrophysiological experiments. All animals were individually housed in a temperature-controlled (24°C) colony room on a 12-hour light/dark cycle. Subjects had ad libitum access to food and water prior to behavioral training. Animals weights were monitored daily when access to water was restricted (see 2.2. Water Restriction), with supplements given as necessary to maintain weight compared to ad libitum littermates. All procedures were approved and

conducted according to guidelines by the Institutional Animal Care and Use Committee (IACUC) at Rutgers, The State University of New Jersey.

2.2.Water Restriction

The behavioral experiments used operant conditioning paradigms to train rats to associate acoustic features of auditory stimuli with water rewards. To motivate animals to perform the instrumental tasks for water reward, rats were placed on a schedule of restricted access to water until they were at 85-90% of their ad libitum water access body weight. Prior to the start of water restriction, rats are weighed daily for at least 3 days to establish individual baseline weights comparable to littermate control animals. While on a schedule of restricted water access thereafter, rats were weighed daily and were given water supplements in their home cage as needed to maintain 85-90% of their baseline weight throughout the duration of the experiments (throughout all training and memory testing).

2.3.Behavioral Apparatus and Stimuli

All behavioral sessions were conducted in two identical instrumental conditioning chambers (H10-11R-TC; Coulbourn Instruments, Holliston, MA) within a sound- attenuated box. Daily training sessions were counterbalanced to ensure equal exposure to both chambers. Each chamber (12” W x 10” D x 12” H; wire mesh floor, H10-11R-TC- NSF) was fitted with a response lever (H21-03R), house light (H11-01R), infrared lights (H27-91R), a speaker (H12-01R), and a water delivery system (H14-05R). During training phases, depressing the response lever (“barpressing”) triggered the presentation of a water cup (~0.02cc) in the reward port (1.25” W x1.625” H). In early barpress (BP) shaping

sessions, a hand switch (H21-01) was also used to trigger presentations of the water cup. Behavioral responses were recorded using Graphic State 4 software (Coulbourn Instruments, Holliston MA) for offline analysis.
All auditory stimuli were generated using Tucker-Davis Technologies (TDT, Alachua, FL) and RPvdsEx software, and presented via the operant chamber’s wall- mounted speaker. White noise (during early instrumental training only) was presented for 7 or 9 seconds in duration (75 dB SPL). Pure tones were always presented for 8 seconds (70 dB SPL). All tone frequencies were cosine-squared gated with rise/fall (10-90%) of 20ms. Sound levels were calibrated daily using a digital sound meter (Larson Davis SoundTrack LxT1).

2.4.Behavioral Training

2.4.1.Initial handling and barpress shaping. After 1 day of acclimating to the vivarium, rats were handled daily for a minimum of 3 days to familiarize them to the transportation to the laboratory and handling during the daily weighing procedure. Two to three days prior to beginning behavioral training, rats were placed on a schedule of restricted water until they reached 85% of their non-restricted weight.
Water-restricted rats were then trained inside the behavioral sound-attenuated chambers to manipulate the wall-mounted lever for water reward to achieve barpress shaping. Barpresses (BPs) resulted in the availability of water reward on a 1:1 ratio. Rewards were presented either for 5 seconds during the first two sessions (which were always on consecutive days), then for 3 seconds on all subsequent daily sessions. The first BP shaping session lasted ~90 minutes or until the animal reached satiety. Remaining

sessions were limited to 45 minutes each. All subjects learned to barpress for water reward within 1-2 sessions. Rats were weighed before and after each session, and a substantial increase in pre- vs. post-training weight (usually near 10.0g per session) immediately confirmed that animals had successfully performed barpresses and obtained water rewards.

2.4.2.Sound control training. After 5 days of BP shaping, animals’ barpressing behavior was placed under sound control. To do so, animals were trained to barpress only during presentations of an auditory stimulus (S) to obtain water reward. BPs made in the presence of the new S (1-12.5kHz band-pass filtered white noise; 75 dB SPL) resulted in the availability of water reward. Barpresses made during the silent inter-trial intervals (ITI; mean:15s, range: 5-25s, randomized) resulted in an error signal (flashing house light) and a “time-out” (an additional 6s lengthening of the ITI until then next trial). To prevent the animal from making associations between the duration of the white noise and reward, the duration of the white noise stimulus varied (either 7 or 9s, 50/50 randomized). Animals could receive a maximum of 2 or 3 rewards per trial (Fig. 1B).
Performance was monitored on a daily basis and calculated separately for each session as follows: Performance = 100%*(# BPs to white noise / Total # BPs). Criterion for moving onto the next stage of training was defined as 2 consecutive days of performance >90% or with an asymptotic level of performance (coefficient of variation = standard deviation/mean, c.v. 0.1) for at least 3 consecutive days. Subjects were trained daily (5 days a week) in single 45-minute sessions until performance criterion was attained. On average, 9 training sessions were required (M = 8.67 days, SD = 1.99 days). All animals

(n = 24) successfully learned to associate the sound with reward, which confirmed that barpressing behavior was under sound control prior to the next phase of training.

2.4.3.Two-tone frequency discrimination task (2TD). To examine whether HDAC3- inhibition during consolidation of auditory associative learning for behaviorally-relevant sounds enhances the specificity of memory formed, we trained rats on a two-tone frequency discrimination task (2TD). After establishing sound control of barpressing behavior, the same water-restricted rats (n = 24) were trained to discriminate between two spectrally distinct sound-frequencies. Barpresses to the S+ (5.0 kHz pure tone) resulted in the presentation of water reward, while barpresses to the S- (11.5 kHz pure tone) were unreinforced and triggered an error signal (flashing house light) and a “time-out” (an extended 6 second interval until the start of the next trial). S+ and S- trials were randomized, and all tones were presented at 70 dB SPL for 8 seconds. Inter-trial intervals (ITIs) were on average 15 seconds (range: 5-25 seconds, randomized). Barpresses during an ITI were inconsequential (i.e., no time-out, no reward). However, barpresses to the S- had a 70% chance of causing a time-out, by extending the following ITI by 6 seconds (Fig. 1A). Sessions were on average 50 minutes in length.
All animals (n = 24) eventually acquired similar levels of 2TD task performance. Daily 2TD performance was calculated as: number of barpresses to the S+ divided by the number of barpresses to both the S+ and S- (P = 100% x [#BPS+/(#BPS+ + #BPS-)]). Performance criterion was defined as 2 consecutive days of performance 90% or 3 days of asymptotic-level performance (c.v. 0.1). Upon reaching performance criterion, rats were trained for 2-3 additional days to insure stable performance prior to memory testing.

Performance of the 2TD task was also assessed using a latency measure to determine the time to the first barpress after the onset of each S sound. Trials in which a barpress was not made were not included in the mean latency to first barpress calculated for each training session. There appeared to be differences in the latencies to barpress to the S+ or S- tones between groups. Thus, an additional performance measure was generated to reflect short-latency barpresses: number of barpresses made during the first 4 seconds to the S+ divided by the number of barpresses to both the S+ and S- during the first 4 seconds [P2 = 100% x latBPS+/(latBPS+ + latBPS-)].

2.4.4.Behavioral assay for associative memory. Following successful 2TD task acquisition and 24 hours after the last 2TD training session, all rats underwent behavioral memory testing to determine effects of HDAC3-inhibition on memory for specific sound frequency. Memory for the acoustic frequency of the associative sound-signals was assessed through a stimulus generalization test (SGT) across acoustic frequency (Fig. 1C).

2.4.5.Stimulus Generalization Test (SGT). To determine memory’s specificity for the tone cues learned in the 2TD task, animals (n = 18/24; n = 9/9 per group) were presented with a range of sound frequencies, including the S+ and S- frequencies (3 animals from each group were used in another analysis not shown). 10 different tone frequencies were tested: 3.6, 4.2, 5.0 (S+), 5.9, 7.0, 8.3, 9.7, 11.5 (S-), 13.6, and 16.0 kHz (all pure tones presented at a sound level of 70 dB SPL). Sound level was the same as used in training. Test tones neighbored the S+ or S- at a distance of ~1/4 octave. The SGT session began with 15 regular trials of the 2TD task (BP to the S+ are rewarded; BPs to the S- are not) to

confirm stable performance on the day of the memory test. Test frequencies were then presented in pseudorandom order over 120 unrewarded trials (to yield 12 presentations of each test frequency; Fig. 1C). The number and latency of behavioral responses to test tones were used to determine the shape of frequency generalization gradients around the two CS frequencies.
The number of BPs made during the SGT for each test frequency were expressed as a proportion of the total number of BPs made for an individual animal, and then averaged for each test frequency to quantify a group mean for the HDAC3i (n = 9) and VEH (n = 9) groups. Latency was measured as the seconds to first BP after tone onset. Frequency- specificity of memory was assessed using the shape of the behavioral gradient across frequencies separately for the excitatory (S+) and the inhibitory (S-) associative memory. Gradient shapes and peaks could determine behavioral sensory acuity and precision, respectively, for each S sound frequency. Thus, nonlinear quadratic regression models were used to approximate the goodness of fit of the actual behavioral gradients to idealized model gradients of highly precise and accurate S memory.

2.5.Electrophysiological Recording

2.5.1.Recording procedures. To determine changes in the frequency-specific responsiveness of auditory cortical tuning, electrophysiological recordings were conducted in an acute and terminal recording session 24-48 hours after the final behavioral SGT. Methods were the same as previously described (Bieszczad et al., 2015 [6]; Bieszczad et al., 2012 [21]). Subjects were anesthetized (sodium pentobarbital, 45mg/kg, i.p.) and administered atropine methyl nitrate (10 mg/kg) to minimize bronchial secretions.

Breathing and movement were monitored throughout the session, and supplemental doses of sodium pentobarbital were given to maintain depth of anesthesia. Ophthalmic ointment was applied to keep the eyes moist and a heating pad was used to maintain core body temperature. Subjects were placed in a stereotaxic frame (Kopf Instruments, Tujunga, CA) inside a double-walled sound attenuated room (Industrial Acoustics Co., Queens, NY). The skull was fixed to a head holder using several small screws and dental acrylic, allowing the removal of ear bars while maintaining the head in a fixed position during auditory stimulation and recording. A craniotomy was performed above the right auditory cortex and the cisterna magna was drained of cerebrospinal fluid to reduce cerebral swelling. The dura over the temporal cortical surface was reflected and warm saline applied intermittently throughout the recording session to prevent desiccation.
Extracellular recordings were made with a linear array (1×4) of parylene-coated microelectrodes (1–2 MΩ, spaced 250 μm apart; FHC, Bowdoin, ME) lowered to the middle cortical layers (III-IV; 400-600 μm depth perpendicular to the cortical surface) via a microdrive (IVM-1000; Scientifica, UK). Electrodes were placed across the auditory cortex (for a total of ~32-160 recording sites per animal). Photographs of the cortical surface were taken to record the placement of each microelectrode and used to create a relative map of the recording locations, which was used to determine subsequent electrode placements for complete coverage of the primary auditory cortical tonotopic surface. While complete A1 maps were not obtained in every animal tested and recorded to construct areal maps of representation in these experiments (e.g., as in Bieszczad et al., 2015 [6]), samples of tuning characteristics across the entire span of caudal-to-rostral A1 representation of

frequency were obtained in each animal to permit a group-based analysis of frequency responsiveness.
Neural activity was amplified x1000, digitized (25 kHz sampling rate; PZ5 NeuroDigitizer Amplifier, RZ6 Bioamp Processor; TDT) and stored for offline analysis. Offline spike detection was performed using custom MATLAB software. Recordings were bandpass filtered (0.3-3.0 kHz). Multiunit discharges were characterized using previously reported temporal and amplitude criteria (Elias et al., 2015 [29]). Acceptable spikes for subsequent analyses of auditory cortical responsiveness were defined as waveforms with peaks separated by no more than 0.6 ms and with a threshold amplitude greater than 1.5 (for the positive peak) and less than 2.0 (for the negative peak) × RMS of 500 random traces from the same recording on the same microelectrode for each site.

2.5.2.Acoustic stimuli Acoustic stimuli were delivered via a calibrated electromagnetic speaker (MF1; TDT Inc.) placed 10-15 cm from the contralateral (left) ear. Pure tone bursts (50 ms, cosine-squared gate with rise/fall time (10-90%) of 10 ms) were presented in pseudorandom order (0.5 – 54.0 kHz in quarter-octave steps; 0-70 dB SPL in 10 dB steps; 5 repetitions). Stimuli were presented with a variable inter-stimulus interval with an average of 700 ± 100 ms apart.

2.5.3.Neurophysiological analysis of bandwidth tuning Frequency response areas (FRAs) were constructed for each recording site using custom MATLAB scripts. Tone-evoked activity was determined by subtracting spontaneous spiking (the 40 ms window prior to tone onset) from evoked-spike activity within a 40 ms response-onset window (6-46 ms

after each tone onset). All responses were determined using spikes/second, i.e., rate. Responses greater than ±1.0 SEM. of the spontaneous spike rate were considered true evoked-responses. Constructed FRAs showed the mean evoked activity to each frequency/sound level combination (168 total) and were used to determine the characteristic frequency (CF) and tuning bandwidth (BW) at each recording site. The borders of each FRA were determined based upon a threshold firing rate value determined by its recorded spontaneous activity; only evoked responses greater than the mean of pre- onset spontaneous activity were considered true sound-evoked responses. Thus, the largest contour was selected from the outlined FRA and was used in determination of the characteristic frequency (CF) for each site. CF was defined as the stimulus frequency having the lowest sound level to evoke a response. If multiple frequencies were found having the same lowest threshold, the CF was defined as the geometric mean between these frequencies, at a frequency that was subsequently verified to have evoked activity. This same FRA contour was used to determine the bandwidth of response at 10, 20, 30 and 40 dB SPL above the evoked sound level threshold for the FRA determined individually at each recording site (see Fig. 7A). Bandwidth was defined as the octave distance between frequencies at the edges of the FRA contour. All recording sites that showed a tone-evoked response consistent with a rostral-to-caudal progression of increasing CFs characteristic of primary auditory cortex were included in the analyses.
Tuning bandwidth was determined above threshold as permitted by the maximum sound level (70 dB SPL) of the acoustic stimuli presented (e.g. if a particular site had a CF threshold at 50 dB SPL, it would preclude analysis of tuning bandwidth at 30 (BW30) and 40 (BW40) dB SPL above the CF; thus only BW10, and BW20 would be determined).

Bandwidths were measured in octave distances and grouped by the CF of each site within a 1/3 octave range (3.4-4.8, 4.8-6.8, 6.8-9.7, 9.7-13.6, and 14.6-19.3 kHz). Therefore, a group mean bandwidth was generated, defined as the average bandwidth within each octave-band (± SEM). Mean CF thresholds and bandwidths were compared between HDAC3i- and VEH-treated groups at each of the BW10, BW20, BW30 and BW40 measurements.

2.6.Pharmacological Inhibition of HDAC3

A pharmacological class I HDAC inhibitor with enhanced selectivity for HDAC3, RGFP966 (10 mg/kg; s.c.), was used to manipulate levels of gene expression. Systemic injections of RGFP966 have been shown to penetrate the blood-brain barrier in rodents (Malvaez et al., 2013 [25]; Bieszczad et al., 2015 [6]; Bowers et al., 2015 [28]) and are effective in promoting histone acetylation linked to gene expression in the rat A1 at this dose, which establishes rationale for a single-inhibitor, single-dose approach. Systemic RGPF966 (10mg/kg) reaches the auditory cortex (Cmax) 30 to 80 minutes post-injection and remains high for at least 4 hours (Bieszczad et al., 2015 [6]). Therefore, systemic RGFP966 delivery at a dosage of 10 mg/kg immediately post-training can reveal effects of HDAC3-inhibition on memory consolidation processes in A1 with subsequent effects on behavioral indices of auditory learning and memory.
Treatment began with 2TD training. Within each performance-matched pair of rats (based on the acquisition of sound controlled barpressing behavior), rats were assigned to receive either RGPF966 (HDAC3i; n = 12) or vehicle (VEH; n = 12). Treatment was counterbalanced across performance: in one half of the paired rats, if there was any slight difference in performance, the higher performing sound-controlled animal received

RGFP966 and the slightly lower performing animal received vehicle; In the other half of the paired rats, the treatment assignment was reversed.
Injections were given immediately after the end of each 2TD training session. Animals received 3 consecutive days of post-session RGFP966 (10mg/kg, s.c.) or VEH injections (of an equivalent volume). All other 2TD sessions were with post-session saline injections, which controlled for any potential effects of injection itself on the rates of acquisition, performance, and subsequent memory. Therefore, any effects on learning and memory could be attributed to the effects of the HDAC3 inhibitor alone.

2.6.Statistical Analyses

All statistical analyses were executed using protocols in Prism 7 software (GraphPad).

2.6.1.Behavior. Differences between HDAC3i and VEH groups were analyzed using ANOVA (=0.05) and independent samples t-tests. Analyses across training sessions (for sound control training and 2TD task acquisition) were performed using 2-way repeated measures ANOVAs, with treatment (“group”) and session as the between-subjects factors. The Bonferroni procedure was used to correct for multiple comparisons in post-hoc tests. SGT data were analyzed separately for the S+ and S- frequencies (also using ANOVA for comparison of the 5 of total 10 frequencies in each S+ or S- comparison). One-sample t- tests (Bonferroni-Holm corrected for multiple comparisons) were used to compare responses that reveal highly frequency-specific responding from generalized responding. Note that multiple comparisons are useful in detecting highly specific (acoustic frequency) effects, which are unlikely to pass ANOVA-based statistical models (across many frequencies). Two-way ANOVAs were used to analyze more global differences between

HDAC3i and VEH groups in responding to test frequencies neighboring the 2TD training tones. Generalization gradients for both S+ and S- tones were compared to idealized generalization gradients of frequency-specific memory, modeled as a non-linear quadratic function centered at the S tone-frequency. Nonlinear regression analysis established goodness of fit (R2 values to the predicted curves). To determine group differences in an SGT shape and peak, the goodness of fit of the dataset to a single curve was compared to goodness of fit to separated group datasets with individually fit curves (assessed by R2 values to the predicted curves and p-values for significance,  = 0.05). This tested for whether the shapes and peaks (or nadirs) of the SGTs in the HDAC3i group were significantly different from the characteristics of the SGTs in the VEH group. Separate analyses were performed for frequency generalization gradients around the S+ or the S- frequency.

2.6.2.Electrophysiology. Independent-samples t-tests were used to compare group differences in CF threshold and tuning bandwidth at 10, 20, 30, and 40 dB SPL above threshold. All tests were corrected for multiple comparisons using the Bonferroni-Holm correction.

3.Results

3.1.HDAC3-inhibition by RGFP966 facilitates associative sound discrimination learning.
To test the hypothesis that HDAC3 inhibition has a key function in the specificity of sensory encoding, we used the class I HDAC inhibitor, RGFP966, to treat rats learning a sound discrimination task. A total of 24 animals (HDAC3i, n = 12; VEH, n = 12) were

trained to asymptotic high levels of performance on a 2-tone discrimination (2TD) task. Successful performance on the 2TD task requires the animal to learn and remember two sound-frequency associations and instrumental responses: barpresses (BPs) to the S+ (5.0 kHz, 70 dB SPL) result in reward delivery, while BPs to the S- (11.5 kHz; 70 dB SPL) do not result in reward and instead trigger an error signal (flashing house light) and a time-out period that extends the time until the next trial. Sound discrimination performance was calculated using the proportion of BPs to the S+ tone relative to BPs to either tone. Barpresses during the silent ITIs are excluded from this performance metric to attribute high performance values to successful sound-frequency discriminations. It is important to note that the frequencies of the two S tones are easily distinguishable to rodents so that performance in the task does not tax perceptual discriminability per se. Instead, increases in the daily performance value is a metric of the animal’s ability to discriminate the two associations for the excitatory S+ frequency and the inhibitory S- frequency to instruct the appropriate associated behavior for obtaining rewards. Performance (%) was calculated as P: