The Thalamocortical Circuit of Auditory Mismatch Negativity

Journal Pre-proof The thalamocortical circuit of auditory mismatch negativity Peter Lakatos, Monica N. O’Connell, Annamaria Barczak, Tammy McGinnis, Samuel Neymotin, Charles E. Schroeder, John F. Smiley, Daniel C. Javitt PII:

S0006-3223(19)31851-7

DOI:

https://doi.org/10.1016/j.biopsych.2019.10.029

Reference:

BPS 14045

To appear in:

Biological Psychiatry

Received Date: 17 July 2019 Revised Date:

30 October 2019

Accepted Date: 30 October 2019

Please cite this article as: Lakatos P., O’Connell M.N., Barczak A., McGinnis T., Neymotin S., Schroeder C.E., Smiley J.F. & Javitt D.C., The thalamocortical circuit of auditory mismatch negativity, Biological Psychiatry (2019), doi: https://doi.org/10.1016/j.biopsych.2019.10.029. This is a PDF file of an article that has undergone enhancements after acceptance, such as the addition of a cover page and metadata, and formatting for readability, but it is not yet the definitive version of record. This version will undergo additional copyediting, typesetting and review before it is published in its final form, but we are providing this version to give early visibility of the article. 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. © 2019 Published by Elsevier Inc on behalf of Society of Biological Psychiatry.

1

The thalamocortical circuit of auditory mismatch negativity. Peter Lakatos1,2, Monica N. O’Connell1, Annamaria Barczak1, Tammy McGinnis1, Samuel Neymotin1, Charles E. Schroeder1,3, John F. Smiley1,2, Daniel C. Javitt1,3 1

Translational Neuroscience Division, Nathan Kline Institute for Psychiatric Research,

Orangeburg, NY, 10962 USA 2

Department of Psychiatry, New York University School of Medicine, NY, 10016 USA

3

Department of Psychiatry, Columbia University College of Physicians and Surgeons, NY,

10032 USA

Corresponding author (lead contact): Peter Lakatos Nathan Kline Institute 140 Old Orangeburg Road, Orangeburg, NY 10962, USA Phone : +1 845 398 6540 Email: [email protected]

Abbreviated title: Thalamocortical mechanism of mismatch negativity

Keywords: Auditory, Thalamocortical, Mismatch Negativity, Non-human Primate, Ketamine, Layer 1

2

ABSTRACT Background: Mismatch negativity (MMN) is an extensively validated biomarker of cognitive function across both normative and clinical populations and has previously been localized to supratemporal auditory cortex. MMN is thought to represent a comparison of the features of the present stimulus vs. a mnemonic template formed by the prior stimuli. Methods: We utilize concurrent thalamic and primary auditory cortical (A1) laminar recordings in 7 macaques to evaluate the relative contributions of core (lemniscal) and matrix (nonlemniscal) thalamic afferents to MMN generation. Results: We demonstrate that 1) deviance-related activity is observed mainly in matrix regions of auditory thalamus; 2) MMN generators are most prominent in layer 1 of cortex as opposed to sensory responses that activate layer 4 first and sequentially all cortical layers; 3) MMN is elicited independent of the frequency tuning of A1 neuronal ensembles. Consistent with prior reports, MMN related thalamocortical activity was strongly inhibited by ketamine. Conclusions: Taken together, our results demonstrate distinct matrix vs. core thalamocortical circuitry underlying the generation of a higher order brain response (MMN) vs. sensory responses.

3

INTRODUCTION Mismatch negativity (MMN) is one of the most extensively studied event-related potential (ERP) measures in both cognitive and applied neuroscience research (1–7). MMN is elicited when a sequence of repetitive stimuli is interrupted by a stimulus that deviates in any of several physical or conceptual dimensions, including frequency, duration, intensity, direction or omission. Thus, because MMN depends upon a comparison between stimuli rather than just a response to stimuli in isolation, it may be considered a perceptual, rather than a purely sensory response. One critical question involving MMN concerns the underlying neural circuits. Generators for auditory MMN have been localized primarily to supratemporal auditory cortex in both humans and non-human primates (8–14) and thus overlap those of sensory components such as auditory N1. Nevertheless, the inputs to cortex that elicit MMN, as well as the nature of the underlying local circuitry remain to be determined. One potential basis for the differential response patterns between perceptual and sensory components comes from consideration of the functional anatomy of the main thalamic auditory relay (medial geniculate body, MGB), which provides predominant input into auditory cortex. Specifically, MGB has at least two discrete thalamocortical projection systems (15, 16). The core (lemniscal) projection system originates largely in the ventral nucleus of the MGB and faithfully transmits stimulus specific information and targets layer 4 of A1. This system shows strong frequency tuning and mapping (tonotopic organization). By contrast, the matrix (“non-lemniscal”) projection system dominates the dorsal and medial nuclei of the MGB. This system targets supra- and infra-granular layers of auditory cortex, shows broader frequency tuning with little if any tonotopic organization. It also projects much wider than the core projection system, targeting even to non-auditory brain regions (15–20), indicating that the matrix may be involved in higher order perceptual-cognitive processes such as attention. Like the thalamic matrix, the pulvinar nucleus of the thalamus is increasingly implicated in higher level neurocognitive processing (21–30). Therefore, for the present study, we recorded from MGB and pulvinar thalamic regions concurrent to all layers of primary auditory cortex in order to evaluate potential subcortical as well as cortical contributions to MMN generation. Along with answering questions about circuit-

4

level mechanisms potentially underlying MMN generation, we also attempted to address issues of stimulus-specific adaptation (SSA) vs. MMN generation in in awake, behaving monkeys. Since MMN is defined as the difference in amplitude between a deviant and standard stimulus, one potential mechanism underlying MMN generation could be a reduction in brain response to a repetitive stimulus (i.e. SSA) (31–41). To date, most prior intracranial studies of MMN have used standard tones whose frequency closely matched the preferred frequency (BF) of the recording site (31, 39, 14), but see (42). We argued that if deviant stimulus related activity occurs in regions that produce little or no evoked type, feedforward response to either the standard or the deviant tones presented, this cannot be simply a result of SSA, since there is no population level response to adapt. This does not mean that certain neuronal populations, like specific inhibitory neuron classes do not increase their firing in response in A1 regions that fall outside the “lemniscal projection zone” related to a certain tone, but this would not constitute a typical, evoked type response measurable in the neuronal ensemble activity. MMN generation has been shown to be particularly sensitive to N-methyl-D-aspartate receptor (NMDAR) antagonists such as ketamine or PCP across human (43–45), monkey (8, 13), and rodent studies (46–51). Moreover, in rodents, effects of PCP may be prevented by simultaneous treatment with NMDAR agonists such as D-serine, similar to observations in schizophrenia (52, 51). Nevertheless, local circuit mechanisms have been studied to only a limited degree. Therefore, in the present study we also assessed the effect of ketamine on both cortical and subcortical deviant stimulus related responses. Our overall results indicate that the effects of deviance detection and SSA are indeed separable and are produced by distinct thalamocortical circuits: the former involving nonlemniscal, thalamocortical matrix projections, and the latter the lemniscal, core thalamocortical system. NMDAR antagonist effects were observed at both subcortical and cortical sites, supporting the importance of NMDAR mediated transmission in the thalamocortical circuit of MMN generation.

5

METHODS Subjects. We analysed the electrophysiological data recorded during 108 A1 and 82 thalamic penetrations using linear array multielectrode in 7 macaque monkeys (4 female and 3 male) who had been prepared surgically for chronic awake recordings. For details of the surgery see Supplementary Methods. Electrophysiology. Neuroelectric activity was obtained using linear array multi-contact electrodes (23 contacts, 100 µm intercontact spacing). The multielectrodes were inserted acutely through guide tube grid inserts, lowered through the dura into the brain, and positioned such that the electrode channels would span all layers of the cortex (first electrode array), or the most active (most auditory responsive) region of MGB/pulvinar (second electrode array), as determined based on inspection of response profiles to 70 dB, 100 ms duration binaural broadband noise bursts in each experiment (Fig. 1A). The angle of all thalamic penetrations corresponded to the angle of cortical penetration and was vertical. In some cases, this allowed for bracketing both MGBd (upper recording channels) and MGBv (lower recording channels) with the 2.3 mm long 23 channel electrode array, as illustrated in figure 1A. The recorded neuroelectric signal was divided into field potential (0.1-300 Hz) and MUA (300-6000 Hz) range. MUA data was also rectified in order to improve the estimation of firing of the local neuronal ensemble. One-dimensional current source density (CSD) profiles were calculated from the local field potential profiles using a three-point formula for the calculation of the second spatial derivative of voltage (53). The advantage of CSD profiles is that they are not affected by volume conduction like the local field potentials, and they also provide a more direct index of the location, direction, and density of the net transmembrane current flow. At the beginning of each experimental session, after refining the electrode position in the neocortex, we established the best frequency (BF) of the A1 and thalamic recording site using a “suprathreshold” method (54–56). To identify the different thalamic subdivisions described in the study, we used a combined anatomical/physiological/machine learning approach. At the end of each animal’s experimental participation, functional assignment of the thalamic recording sites is confirmed histologically by the reconstruction of a subset of electrode tracks through post-mortem histology

6

(Nissl staining, parvalbumin and calbindin immunohistochemistry, e.g. Figure 1B). For further details see Supplementary Methods. Behavioral task and stimuli. Subjects were awake during the recordings, but they were not required to respond to the stimuli presented. The data presented here were recorded during trial blocks in which non-overlapping visual and auditory stimuli were presented simultaneously. Only auditory stimulus related activity was analyzed. The auditory stimulus stream consisted of pure tone beeps with a constant SOA of 624.5-ms. The frequency of the auditory standards was varied across stimulus blocks within the monkey’s hearing range (0.3-32 kHz). Frequency deviants (2-4 semitones different from the standard) occurred in the stream of standard tones every 3-9 seconds randomly. Maximally, 14 different frequency tone streams were presented per experiment, but after rejecting trial blocks with considerable movement artifacts (affecting more than 20% of deviant tone related responses), we were left with 1-13 trial blocks per recording site presented (mean=3, std=2.5) (see also Supplementary Methods. To examine the effect of ketamine on deviant-related thalamocortical responses and surface MMN, in a subset of experiments (12 A1 and 8 MGBd/m sites in 2 monkeys), we administered systemic subanesthetic doses of ketamine (1 mg/kg i.m.). After recording a control trial block, we injected the ketamine in the gluteus maximums of the subjects without removing them from the chair and recorded for 55 minutes at 10-minute intervals (6 post-ketamine trial blocks). Data analyses. Data were analyzed offline using native and custom-written functions in Matlab (Mathworks, Natick, MA). When comparing MUA and CSD responses to standard and deviant tones, we analyzed the same number of responses: responses to all deviant stimuli and responses to the standard that occurred just prior to each deviant (Figures 2-4). For time-frequency analyses (Figures 3-4), continuous oscillatory amplitudes and phases were extracted by wavelet decomposition (see Supplementary methods).

7

RESULTS We used linear array, 23 contact multielectrodes to record neuroelectric activity from 108 A1 and 82 thalamic sites (MGB ventral nucleus (MGBv, n=18), MGB dorsal and medial nucleus (MGBd/m, n=42), and pulvinar (n=22)) in 7 macaque monkeys (Fig. 1A). All analyses were performed on CSD and MUA signals, reflecting transmembrane currents and firing of the neuronal ensemble recorded from respectively.

Frequency specificity of stimulus specific adaptation in A1 We analyzed deviant-related thalamocortical activity using a classic auditory oddball paradigm with infrequent (~10% sequential probability) frequency “deviants” embedded in a rhythmic stream of frequently occurring “standard” pure tones. Frequency deviants (2-4 semitone difference) occurred randomly at 3 to 9 second intervals. Most previous rodent and non-human primate studies of deviant-related activity have utilized stimulus streams with puretone frequency that was a close match to the BF of the recording site (8, 31, 39, 14), we also presented streams with tone frequencies as much as 6 octaves higher or lower than the BF of the A1 recording site. Our goal was to test the hypothesis that the deviant- related activity associated with MMN occurs in regions that lack lemniscal thalamocortical inputs related to the tones presented. Figure 1B shows A1 (upper) and auditory thalamus (lower) MUA responses for the first 7 tones in auditory streams, averaged across all trial blocks. Stars denote significant response amplitude differences compared to the first response in the 5-30 ms post-stimulus timeframe for each of the 6 responses following the first response, since if stimulus specific adaption was to occur, these subsequent responses would have significantly lower amplitudes (31, 57, 32, 34, 38). Applying the 5/3 rule for extrapolating from human to monkey sensory response latencies (58), the 5-30 ms timeframe corresponds to the “middle-latency response” timeframe in which correlates of SSA occur in humans (10-50 ms) (59). As Fig. 1B shows, SSA indeed occurs in A1, but is not detectable in the thalamus, using the same methods. Next, we examined the same responses in A1 in 4 separate groups that were binned based on their best frequency – tone frequency (BF-TF) difference, from 0 difference, which

8

corresponds to matching BF and TF, to 6 octaves difference (Fig. 1C histogram). We found that the adaptation of MUA responses to frequently presented standard tones – i.e. lower response amplitudes following the first response – is detectable only in A1 sites that are closely tuned to the frequency of tones presented (< 1 octave BF-TF difference, Fig 1C, top trace). This indicates that at least on the neuronal ensemble level, SSA only occurs in A1 regions receiving lemniscal thalamocortical inputs related to the auditory stimuli presented, resulting in an evoked type response.

Deviant-related A1 MUA response is not frequency specific Figure 2A and C show that overall, both A1 and Thalamic MUA responses to deviant tones are larger than to the standards immediately preceding the deviants. Next, we investigated deviant stimulus related MUA responses of A1 in the same “frequency specific” manner as SSA (in 4 groups binned on BF-TF difference). We reasoned that if deviant-related activity simply reflects SSA, it should not be detectable at A1 sites that are not tuned to respond to the tones presented, and therefore, do not produce an evoked type response. We found that in contrast to SSA, deviant-related MUA increase was significant for all 4 A1 site groups, independent of BF-TF difference (Fig. 2B). We also noted that deviant-related MUA response in the early, 5-25 ms timeframe, in which SSA occurs (57, 39, 59), was only significantly greater than standard tone related response for the blocks in which BF and TF were a close match (< 1 octave difference) (Fig. 2B top). Therefore, for these blocks, SSA and deviant-related activity likely co-occur and might be inseparable as previously suggested (39). Blocks closest to the <1 octave BF-TF condition might still show slight SSA effects due to its proximity to the BF region. We therefore determined the timecourse of deviant-standard MUA response differences for the 2 groups in which the BF-TF difference was >2.5 octave (2 bottom traces in Fig. 2B). The deviant-related response onset in these two groups was a close match (33 vs. 32-ms). To explore a possible thalamic origin of the deviant-related A1 MUA response, we examined standard- and deviant-related response amplitude differences in 3 distinct thalamic regions: the lemniscal nucleus of the MGB, the non-lemniscal nuclei constituting the thalamic matrix, and the pulvinar, which is the largest thalamic nucleus in human and non-human

9

primates (60). While MUA responses of the lemniscal thalamic nucleus (MGBv) to standard vs. deviant stimuli do not differ significantly, both non-lemniscal nuclei of the MGB (MGBd/m) and the pulvinar show a significant deviant-related response (for the 25-50 ms time interval: pMGBv=0.112; pMGBd/m=0.0014; ppulvinar=0.0079, Wilcoxon signed rank, Bonferroni corrected) (Figure 2C). Furthermore, in the MGBd/m nuclei the deviant-related response increase occurs earlier (24 ms) and ceases around 50 ms, in the pulvinar this response augmentation arises later (34 ms) and appears to last for at least up to 200 ms post-stimulus. To quantify this, we performed a 5ms non-overlapping sliding window analysis of the standard vs. deviant MUA responses. We found that the first significant difference occurred in the MGBd/m data in the 21-25 ms window (p=0.0017, Wilcoxon signed rank, Bonferroni corrected). The difference between standard vs. deviant MUA responses first occurred in the 31-35 ms window for pulvinar (p=0.049, Wilcoxon signed rank, Bonferroni corrected), but never reached a significant level in any time windows for MGBv. Furthermore, while MGBd/m deviant-related MUA activity was not significant in any time window after 50 ms, pulvinar deviant-related MUA activity was significantly different even in the 196-200 ms time window, supporting the notion that deviant-related pulvinar activity starts later, and is more sustained than deviant-related activity of non-lemniscal (matrix) MGB. To examine how widespread deviant related thalamic response is, we also determined the percentage of electrode sites with a significant deviant vs. standard stimulus related MUA response enhancement in the 25-50 ms timeframe. Figure 2D displays standard and deviant stimulus related responses for this subset of sites, and the inset shows that the majority are found in the non-lemniscal MGB and pulvinar nuclei, indicating a larger role of the thalamic matrix in deviant stimulus related activity compared to the core. Note that since the terms core vs. matrix and lemniscal vs. non-lemniscal refer to similar thalamocortical projection systems of the MGB (largely MGBv vs. MGBd/m), we use these interchangeably throughout the manuscript, as we feel that these terminologies, based on different anatomical observations (connectivity vs. immunohistochemistry), are overlapping and are equally important when communicating our results.

10

CSD correlate of the surface mismatch negativity (MMN) component in A1 Prior monkey studies have shown deviant-related activity in the surface EEG at ~50-100 ms (61, 13, 62). Here, we also found significant MMN-like activity during 55 – 90 ms timeframe in the surface EEG recordings (n=102, p=0013, Wilcoxon signed rank), with the peak negativity at 75 ms (Figure 3A). To examine the relationship between surface and intracranial activity, we inspected standard and deviant stimulus related laminar current source density profiles and found that as Figure 3B shows, the largest differences occurred in the most superficial recording locations close to the pial surface, as well as some deeper locations, which correspond to the infragranular layers. To quantify this observation, we identified single electrode locations with a significant deviant vs. standard stimulus related sink (negativity, Wilcoxon signed rank, p < 0.05, Bonferroni corrected). Figure 3C shows the traces of deviant and standard stimulus related responses from these locations, which are of different polarities (deviant-related sink vs. standard related source), indicating that the deviant stimulus related response is not merely an amplification of the standard stimulus related response, but rather is qualitatively different. The bar plot to the right designates the distance of the recording sites with a significant deviant-related sink relative to the pial surface of A1 (top black horizontal line), and relative to the top of the granular (4, middle black line) and top of the infragranular (layer 5, bottom black line) layers. The distribution of significant deviant-related sinks is heavily biased to layer 1, indicating that this most superficial layer might play a key role in deviant-related neuronal activity. Specifically, the likely origin of the surface MMN component is a supragranular sink over source pair, with the active sink – signaling excitation – located in layer 1, signaling a reversal of excitability at the apical dendrites of cortical pyramidal neurons when the brain detects a deviant stimulus (Fig. 3D). As Figure 3E shows, the distribution of sites with a significant deviant-related CSD response was even across different BF-TF differences, verifying MUA-based results that deviant-related responses are not frequency-specific in A1. Since a previous macaque study found the most significant SSA in the granular layer of A1 (39), we also examined the granular layer sink-source pair in A1 sites with significant

11

deviant-related supragranular responses. Stunningly, we found no significant difference between standard and deviant-related granular CSD activity (Fig. 3F) in the 25-75 ms time interval (n=196, for sinks: p=0.66, for sources: p=0.227, Wilcoxon signed rank), despite the large significant supragranular response difference (Fig. 3D). This further supports the notion that while the lemniscal, granulo-centric thalamocortical system might play a role in SSA, deviantrelated extragranulo-centric non-lemniscal activity arises via an independent circuit. Next we examined the time-frequency signatures of A1 and surface recordings (Figure 3G), and found that: 1) the largest significant standard- vs. deviant-related difference in surface EEG is a deviant-related increase in post-stimulus intertrial coherence (ITC) within the theta frequency band (6-9Hz), with little associated power change, and 2) a corresponding difference occurs also in supragranular CSD signals, accompanied by smaller changes in higher frequency bands. This supports the notion that surface MMN at least partly reflects supragranular A1 responses to deviant stimuli. It is also apparent from the figures that phase-based time frequency differences between responses to standard vs. deviant tones are much more apparent than amplitude-based differences, possibly indicating phase reset as a mechanism. The reason is that phase rest of neuronal oscillations aligns oscillatory phases to the timing of resetting events without any added neuronal activity or evoked type response. To test this, we compared deviant and standard tone related ITC and oscillatory amplitude in 6 frequency bands (delta1 (0.8-2.5 Hz), delta2 (2.6 -4 Hz), theta (4.5 – 9 Hz), alpha (9.5 – 14 Hz), beta (15 – 27 Hz), gamma (28 – 55 Hz), high gamma (65 – 180 Hz)) in both intracranial and surface recordings. We found that while theta ITC difference was significant both in layer 1 and surface recordings (player1=0.0029; psurface=0.0096, Wilcoxon signed rank, Bonferroni corrected), none of the other variables including theta amplitudes were significantly different in response to standard vs. deviant tones, indicating a specific reset of theta oscillatory activity in response to deviant tones. This does not mean that the only cortical response to deviants is theta phase reset (see the significant deviantrelated amplitude increases in figure 3G, supragranular CSD), but it does show that theta phase reset is a dominant component of deviant detection in the brain.

12

The effect of subanesthetic ketamine on deviant-related thalamocortical responses. To further examine the idea that the deviant-related thalamocortical activity involving the non-lemniscal thalamus and supragranular A1 layers is a correlate of the surface MMN, we administered systemic subanesthetic doses of ketamine (1 mg/kg i.m.) and determined its effect on deviant-related neuronal activity in a subset of experiments (12 A1 and 8 MGBd/m sites in 2 monkeys). Figure 4A shows that, as demonstrated by previous studies (44, 13, 63), the significant deviant-related negativity in response to deviant stimuli (MMN) prior to ketamine injection (left panel, control) virtually disappears following ketamine administration (middle panel, 5 minutes post-ketamine). In intracranial recordings, the deviant-related MUA responses in A1 and MGBd/m became suppressed following ketamine administration (Fig. 4B-C), and the deviant stimulus related sink was significantly reduced following ketamine injection (Fig. 4D). Notably, for both MUA and CSD, significant suppression of deviant-related responses only occurred in the “late” post-stimulus timeframe (25-75 ms for A1, 20-40 ms for MGBd/m, light green traces), and not in the early timeframe (5-25 ms for A1, 3-20 ms for MGBd/m, dark green traces). Post-ketamine supragranular CSD response suppression presented as a diminished theta frequency band ITC difference between standard and deviant stimuli in the time-frequency domain (Fig. 4E). To quantify this, we compared theta ITC (6-9 Hz) pre and 5 minutes postketamine injection in the 25-75 ms timeframe, and found that the significantly larger theta ITC related to deviant stimuli vanished following ketamine administration (Wilcoxon signed rank, n=12, p=0.009 vs. p=0.677).

13

DISCUSSION Our current study confirms and extends prior reports of MMN generation within supragranular layers of primary auditory cortex in non-human primates. We show that the strongest intracortical correlate of the MMN is located in Layer 1, and it mostly involves the reset and alignment of theta frequency band neuronal activity to the onset of deviant stimuli. We also demonstrate that in the thalamus, the non-lemniscal, matrix system likely plays a larger role than the lemniscal MGB core in the elicitation of MMN, as most deviant versus standard stimulus related response enhancements were localized to MGBd/m and the pulvinar. These two major novel findings are in alignment as the thalamic matrix projects mainly to Layer 1, as opposed to the thalamic core which targets mainly Layer 4. To our knowledge, this is the first study to evaluate potential contributions of the thalamic matrix and core to MMN generation, as well as the first to investigate the mechanism – modulatory vs. driving – of the intracortical correlate of MMN. Admittedly, while most of our results are correlational, they allow us to outline a hypothetical deviant detection and activation circuit. We propose that the key constituents of this circuit are the non-lemniscal nuclei of the MGB (MGBd/m), the thalamic reticular nucleus (TRN), layers 1 and 5 of A1, and the pulvinar, which are all elements of the auditory thalamic matrix system (16, 24). The current results show that deviant-related activity in A1 is not frequency specific, that it occurs in regions tuned to as far as 6 octaves apart and has the same “sign” indicative of excitation in layer 1 (occurring likely through disinhibition, see Supplementary Discussion). It is characterized by a reversal to a sink over source response pattern compared to standard stimuli (a source over sink pattern) in the supragranular layers with no change in the response of the granular layer. Thus, deviant-related activity cannot be mediated by lemniscal thalamic inputs from MGBv that are sharply tuned and project to the granular layer of narrow A1 regions (64–67). On the other hand, non-lemniscal or matrix dominated nuclei of the MGB (MGBd/m) exhibit broad frequency tuning, and project broadly to the supragranualar layers (mainly layer 1) of A1 and higher order auditory cortical regions (68, 69, 16), and are likely responsible for the deviant stimulus related response. Indeed, we found that deviant stimulus-related MUA response

14

enhancement mostly occurs in the non-lemniscal (matrix dominated) nuclei of MGB (MGBd/m), starting around 25 ms. Since the onset of deviant vs. standard stimulus related MUA activity differences occurs much later than MUA response onset in MGB at around 3 ms (28), or even response onset in A1 starting at around 5 ms (55), we hypothesize that deviant detection involves a brain region other than the MGB. Previous studies show that the physiological and anatomical properties of the TRN make this structure an ideal candidate. Two of the key properties that potentially endow TRN with deviant detection capabilities are that TRN neurons are highly sensitive to auditory frequency deviants (70), and that its circuitry is ideally suited to suppress frequently repeating inputs and enhance salient ones (71). The role of the TRN is also supported by the study of Yu et al. (70), who demonstrated that TRN neurons responded more strongly to pure-tone deviant vs. standard stimuli than MGB neurons, which also exhibited deviance detection properties but to a lesser extent. Besides pointing to a rudimentary hypothetical circuit, our results also dispute the idea that MMN is merely a correlate of SSA, as they are likely generated by, and their effects are disseminated by, distinct thalamocortical circuits. Most importantly, while previous studies suggested that SSA preferentially occurs in regions that produce an evoked type response to standard and deviant tones via frequency specific lemniscal pathways (31, 34, 37, 39), our results show that the thalamocortical correlates of MMN are not frequency specific. This difference is potentially very important from a functional standpoint. The likely reason it has remained undiscovered thus far is that in most animal studies, the frequency of tones in oddball paradigms is adjusted to be in the vicinity of the best frequency of auditory cortical and thalamic regions studied. We speculate that the practical reason deviant stimulus related thalamocortical neuronal activity identified by our study is non-frequency specific is that the function of such activity is more “alerting” rather than “analyzing”. These two distinct functions might be supported by two distinct dynamical components of auditory short-term memory: a detail rich component with rapid decay and a more stable but less detail-rich component (72). Finally, our findings indicate that the dominant spectral signature of the monkey MMN is a deviant-related intertrial coherence (ITC) increase focused in the theta band, without much amplitude increase in the same or other frequency bands in layer 1 and surface recordings. This indicates that a major mechanism underlying the MMN component is oscillatory phase reset,

15

characterized by narrow frequency “phase locking” (indexed by ITC) related to a certain input – in this case the deviant stimulus related input – without any associated amplitude increase in the same frequency band. This further differentiates MMN from SSA, as SSA clearly operates on “evoked” responses, characterized by stimulus related MUA and CSD amplitude modulation as shown by previous research (31, 34, 37, 39) and our data (Figs 1C & 2B). To summarize, previous studies show that there are two main, functionally distinct types of event related responses in the brain, “stimulus-evoked” and “phase-reset,” whose main spectral signatures are power increase and phase concentration, respectively (73, 67, 74). We propose that these correspond to the “driving” and “modulatory” response types described by Sherman and colleagues (21, 75, 76). While SSA involves mostly the former, MMN is mostly generated by the latter mechanism. The two functionally distinct response types are also indicative of two anatomically distinct thalamocortical circuits involved in their generation: while spatially confined evoked responses in A1 are a consequence of Layer 4 targeting core thalamic projections, modulatory A1 activation occurs via the thalamic matrix that projects broadly to Layer 1 of A1. Upon detection of a deviant stimulus possibly by the TRN matrix projections initiate an A1 wide reset of theta oscillations to their high excitability phases, that has the potential to amplify any deviant stimulus related core thalamocortical inputs.

Acknowledgements This work was supported by NIH grants R01DC012947 to PL and SN, R01MH109289 and R01MH49334 to DCJ, P50MH109429 to PL and CES, and ARO grant W911NF-19-1-0402 to SN. . Conflicts of Interest. The authors report no biomedical financial interests or potential conflicts of interest.

16

FIGURE LEGENDS

Figure 1 Multielectrode recordings and stimulus specific adaptation (SSA) in the auditory thalamocortical system. (A) Schematic of a linear array multielectrode and representative CSD (left) and MUA (right) profiles for recordings in A1 (top) and MGB (bottom). Transmembrane currents [sinks (negative) and sources (positive)] in CSD colormaps are color coded (red and blue, respectively). Horizontal blue dashed lines mark boundaries of supragranular, granular, and infragranular layers in A1, and the boundary between the dorsal and ventral nuclei in MGB. (B) Parvalbumin immunohistochemistry of auditory cortex and calbindin immunohistochemistry of MGB. The arrows mark electrode tracks. (C) 5-second-long MUA traces display the first seven responses to standard tones presented in 302 trial blocks while recording in A1 and 160 trial blocks while recording in Thalamus (MGB and medial pulvinar). Stars denote MUA response difference compared to the first response in the 5-30 ms post-stimulus timeframe (Wilcoxon signed rank, p<0.05, Bonferroni corrected). Note the pronounced SSA in A1 and the lack of response adaptation in the thalamus. (D) The bar graph displays the difference between the Best Frequency (BF) of each A1 recording site and the frequency of standard pure tones presented during A1 recordings. 0 corresponds to when the frequency of standard pure tones presented matched the BF. Color coding signifies the 4 groups for which MUA responses were selectively averaged and plotted below (best frequency – tone frequency (BF-TF)=0-0.5 octave, n=102; 12.5 octave, n=104; 3-4.5 octave, n=79; and 5-6 octave, n=41 from top to bottom). Stars denote MUA response difference compared to the first response in the 5-30 ms post-stimulus timeframe (Wilcoxon signed rank, p<0.05, Bonferroni corrected).

Figure 2 Deviant tone specific multiunit activity responses in the auditory thalamocortical system. (A) Multiunit activity averaged across all A1 layers and experiments in response to rarely occurring frequency deviant (red) and frequently repeated standard (blue) tones (mean frequency difference=20%, standard deviation=10%). Note that significant response amplitude difference (Wilcoxon signed rank, p<0.05, Bonferroni corrected), denoted by the orange strip below the x-axis, occurs almost immediately after response onset, at 16 ms. (B) Frequency

17

dependence of early (5-25 ms) vs. late (>25 ms) MUA responses to deviant tones. The MUA traces display response to standard and deviant tones selectively averaged based on best frequency – tone frequency (BF-TF) difference (same groups as Figure 1C, see magenta labels on top of each graph). Stars denote significant response amplitude differences in the timeframes demarcated by the gray dotted lines: 5-25 ms, 25-75 ms, 75-125 ms, and 125-175 ms (Wilcoxon signed rank, p<0.05, Bonferroni corrected). Note that while the early response amplitude difference is only observable in the first group, when the frequency of presented tones and the BF of the recording site is a close match, the later response amplitude differences are frequency independent and occur in all four groups. (C) Thalamic MUA averaged across all nuclei on top and nucleus specific MUA traces on bottom showing responses to standard and deviant tones. Like the previous panels, the orange strip below the x axis denotes time resolved significant response amplitude differences, while the orange stars denote differences in the 4 different timeframes. Note that while neuronal firing in the ventral nucleus is not increased in response to deviant stimuli, there is a significant response amplitude increase in the 25-75 ms timeframe in the non-lemniscal nuclei of the MGB and the pulvinar (Wilcoxon signed rank, p<0.05, Bonferroni corrected). The pulvinar also displays a long-lasting response amplitude increase to frequency deviants. (D) Thalamic MUA responses to standard and deviant tones averaged across single electrode recording sites (up to 21 sites per recording corresponding to the individual electrodes of the linear electrode array during a trial block) where single trial standard vs. deviant tone related response amplitudes were found to be significantly different in the 25-50 ms timeframe. The bar plot insert shows how these recording sites were distributed across the different thalamic nuclei (% of recording sites with significant deviant tone specific response, n=8 for MGBv, n=138 for MGBd/m, and n=55 for pulvinar).

Figure 3 Surface (EEG) recorded mismatch negativity and underlying transmembrane currents (CSD) in primary auditory cortex. (A) Responses to standard (blue) and deviant (red) tones recorded above the brain by an electrode placed in the recording chamber filled with saline. The orange strip below the x -axis denotes time resolved significant response amplitude differences. The green trace to the left shows the difference waveform (deviantstandard), with the vertical green dotted lines demarcating time periods during which the

18

deviant-related response is significantly more negative across trial blocks. The orange stars signal that the difference waveform averaged in the earlier (56-91 ms) and later (166-191 ms) timeframes is significantly different from 0 (Wilcoxon rank sum, p=0.002 and p=0.020 respectively). (B) Current source density response profiles from two representative A1 sites in response to standard (left) and deviant tones (middle). The colormaps to the right show the difference CSD maps. The BF of the site whose responses are displayed in the top row matched the frequency of tones presented, while the BF of the site on the bottom was 2 octaves different than the tones presented. Note the inversion or near inversion of sources (in standard related responses) to sinks (in deviant-related responses) in higher supragranular locations likely corresponding to layer 1 marked by the white dotted lines in both sites. (C) The red trace shows deviant-related CSD responses from A1 recording sites with a significant deviant vs. standard stimulus related sink within one trial block. The blue trace is the standard related CSD responses from the same sites. Significant response amplitude difference denoted by the orange strip below the x axis occurs from 22 ms on. The bar graph to the right displays the number of recording sites with significant deviant-related sinks in units of 0.1mm relative to the top of the cortex (pia), the top of layer 4, and the top of layer 5, all marked by black horizontal lines. (D) Deviant (left) and standard (right) related sink-source response pairs from supragranular recording sites. The location of the source (passive current) corresponding to the active deviant-related sink in the supragranular layers (traces on the left) was located on average 0.50 millimeters below the sink (SD=0.45 millimeters). As opposed to this, in the same pair of recording locations we found an overall source over sink (reversed) response pattern related to standard stimuli (traces on the right), indicating that standard vs. deviant stimulus related response are opposite in sign in layer 1. (E) Percentage of different Best Frequency – Tone Frequency (BF-TF) trial blocks for which significant deviant-related sink was detected. (F) Granular sink-source response pairs from the same A1 locations. Note that despite the large difference in deviant vs. standard related response amplitudes in supragranular sites, there is no difference in the amplitude of granular responses. (G) Time-frequency properties of standard (left) vs. deviant (middle) tone related responses. The upper 4 colormaps display time-frequency ITC and amplitude maps of the supragranular recording sites with a significant deviant-related sink (same as (D)), while the lower 4 colormaps display time-frequency ITC and amplitude maps of surface recordings. Traces to the far right

19

display the same variables averaged in the timeframe of significant surface mismatch negativity (56 – 91 ms, see panel (A)). The orange strips near the y axes denote a significant difference between standard tone and deviant tone related response properties.

Figure 4 The effect of systemic ketamine on the mismatch effect in the thalamocortical system. (A) Responses to standard (blue) and deviant (red) tones recorded above the brain before subanesthetic (1 mg/kg i.m.) ketamine injections (control) and 5 minutes following drug administration. The orange strip below the x -axis denotes time resolved significant response amplitude differences. The light green trace to the right shows the difference of deviant and standard related responses across all trial blocks in the 65-110 ms timeframe recorded at 0 – 55 minutes following ketamine injection, with 0 marking the control condition. The orange stars mark significant differences from the control recording (Wilcoxon signed rank, p < 0.05, Bonferroni corrected). The BF-TF difference for this subset of recordings ranged from 0 to 5.5 octave (mean=3.1, std=1.3). (B) Standard and deviant tone related A1 MUA responses before subanesthetic ketamine injections (control) and 5 minutes following drug administration. The orange strip below the x axis denotes time resolved significant response amplitude differences. Note that the deviant-related increase in MUA in the 25-75 ms timeframe disappears following ketamine injection (p=0.013 vs. p=0.898 respectively, Wilcoxon signed rank). The green traces to the far right show the ketamine effect on deviant-standard MUA response amplitudes for an early (5-25 ms, dark green) and a late (25-75 ms, light green) timeframe across 6 consecutive post-ketamine injection recordings lasting 1 hour. The orange stars mark significant differences from the control recording. (C) Same as A for non-lemniscal (MGBd/m) thalamic recordings, except that the early timeframe for thalamus is defined as 3-20 ms, while the late timeframe as 20-45 ms. (D) Same as B and C for supragranular locations with a significant deviant-related sink. (E) Time frequency ITC properties of standard (left) and deviant (right) responses prior to (top) and following (bottom) subanesthetic ketamine injections. It is apparent from the colormaps and from the traces to the right displaying ITC averaged in the 25 – 75 ms timeframe that ketamine abolishes the theta ITC hallmark of deviant vs. standard tone related activity. Deviant stimulus related theta ITC was significantly larger than standard stimulus related theta

20

ITC in prior to (Wilcoxon signed rank, n=12, p=0.009), and not different following ketamine administration (Wilcoxon signed rank, n=12, p=0.677). REFERENCES 1.

Näätänen R, Paavilainen P, Rinne T, Alho K (2007): The mismatch negativity (mmn) in basic research of central auditory processing: a review. Clin. Neurophysiol. 118(12): 2544– 2590.

2.

Javitt DC, Spencer KM, Thaker GK, Winterer G, Hajós M (2008): Neurophysiological biomarkers for drug development in schizophrenia. Nat. Rev. Drug Discov. 7(1): 68–83.

3.

Garrido MI, Kilner JM, Stephan KE, Friston KJ (2009): The mismatch negativity: a review of underlying mechanisms. Clin. Neurophysiol. 120(3): 453–463.

4.

Näätänen R, Kujala T, Winkler I (2011): Auditory processing that leads to conscious perception: a unique window to central auditory processing opened by the mismatch negativity and related responses. Psychophysiology. 48(1): 4–22.

5.

Light GA, Näätänen R (2013): Mismatch negativity is a breakthrough biomarker for understanding and treating psychotic disorders. Proc. Natl. Acad. Sci. USA. 110(38): 15175–15176.

6.

Perez VB, Woods SW, Roach BJ, Ford JM, McGlashan TH, Srihari VH, et al. (2014): Automatic auditory processing deficits in schizophrenia and clinical high-risk patients: forecasting psychosis risk with mismatch negativity. Biol. Psychiatry. 75(6): 459–469.

7.

Avissar M, Javitt D (2018): Mismatch negativity: a simple and useful biomarker of nmethyl-d-aspartate receptor (nmdar)-type glutamate dysfunction in schizophrenia. Schizophr. Res. 191: 1–4.

8.

Javitt DC, Steinschneider M, Schroeder CE, Vaughan HG, Arezzo JC (1994): Detection of stimulus deviance within primate primary auditory cortex: intracortical mechanisms of mismatch negativity (mmn) generation. Brain Res. 667(2): 192–200.

9.

Javitt DC (2000): Intracortical mechanisms of mismatch negativity dysfunction in schizophrenia. Audiol Neurootol. 5(3-4): 207–215.

10.

Pincze Z, Lakatos P, Rajkai C, Ulbert I, Karmos G (2001): Separation of mismatch negativity and the n1 wave in the auditory cortex of the cat: a topographic study. Clin. Neurophysiol. 112(5): 778–784.

11.

Liebenthal E, Ellingson ML, Spanaki MV, Prieto TE, Ropella KM, Binder JR (2003): Simultaneous erp and fmri of the auditory cortex in a passive oddball paradigm. Neuroimage. 19(4): 1395–1404.

21

12.

Sussman E, Steinschneider M (2006): Neurophysiological evidence for context-dependent encoding of sensory input in human auditory cortex. Brain Res. 1075(1): 165–174.

13.

Gil-da-Costa R, Stoner GR, Fung R, Albright TD (2013): Nonhuman primate model of schizophrenia using a noninvasive eeg method. Proc. Natl. Acad. Sci. USA. 110(38): 15425–15430.

14.

Chen I-W, Helmchen F, Lütcke H (2015): Specific early and late oddball-evoked responses in excitatory and inhibitory neurons of mouse auditory cortex. J. Neurosci. 35(36): 12560– 12573.

15.

Jones EG (1998): Viewpoint: the core and matrix of thalamic organization. Neuroscience. 85(2): 331–345.

16.

Jones EG (2001): The thalamic matrix and thalamocortical synchrony. Trends Neurosci. 24(10): 595–601.

17.

Shipp S (2007): Structure and function of the cerebral cortex. Curr. Biol. 17(12): R443–9.

18.

Harris KD, Mrsic-Flogel TD (2013): Cortical connectivity and sensory coding. Nature. 503(7474): 51–58.

19.

Allene C, Lourenço J, Bacci A (2015): The neuronal identity bias behind neocortical gabaergic plasticity. Trends Neurosci. 38(9): 524–534.

20.

Harris KD, Shepherd GMG (2015): The neocortical circuit: themes and variations. Nat. Neurosci. 18(2): 170–181.

21.

Sherman SM, Guillery RW (2002): The role of the thalamus in the flow of information to the cortex. Philos. Trans. R. Soc. Lond. B, Biol. Sci. 357(1428): 1695–1708.

22.

Saalmann YB, Pinsk MA, Wang L, Li X, Kastner S (2012): The pulvinar regulates information transmission between cortical areas based on attention demands. Science. 337(6095): 753–756.

23.

Barron DS, Eickhoff SB, Clos M, Fox PT (2015): Human pulvinar functional organization and connectivity. Hum. Brain Mapp. 36(7): 2417–2431.

24.

Lakatos P, O’Connell MN, Barczak A (2016): Pondering the pulvinar. Neuron. 89(1): 5–7.

25.

Roth MM, Dahmen JC, Muir DR, Imhof F, Martini FJ, Hofer SB (2016): Thalamic nuclei convey diverse contextual information to layer 1 of visual cortex. Nat. Neurosci. 19(2): 299–307.

26.

Sherman SM (2016): Thalamus plays a central role in ongoing cortical functioning. Nat. Neurosci. 19(4): 533–541.

27.

Zhou H, Schafer RJ, Desimone R (2016): Pulvinar-cortex interactions in vision and attention. Neuron. 89(1): 209–220.

22

28.

Barczak A, O’Connell MN, McGinnis T, Ross D, Mowery T, Falchier A, et al. (2018): Top-down, contextual entrainment of neuronal oscillations in the auditory thalamocortical circuit. Proc. Natl. Acad. Sci. USA. 115(32): E7605–E7614.

29.

Fiebelkorn IC, Kastner S (2019): The puzzling pulvinar. Neuron. 101(2): 201–203.

30.

Fiebelkorn IC, Pinsk MA, Kastner S (2019): The mediodorsal pulvinar coordinates the macaque fronto-parietal network during rhythmic spatial attention. Nat. Commun. 10(1): 215.

31.

Ulanovsky N, Las L, Nelken I (2003): Processing of low-probability sounds by cortical neurons. Nat. Neurosci. 6(4): 391–398.

32.

Anderson LA, Christianson GB, Linden JF (2009): Stimulus-specific adaptation occurs in the auditory thalamus. J. Neurosci. 29(22): 7359–7363.

33.

Rui Y-Y, He J, Zhai Y-Y, Sun Z-H, Yu X (2018): Frequency-dependent stimulus-specific adaptation and regularity sensitivity in the rat auditory thalamus. Neuroscience. 392: 13– 24.

34.

Szymanski FD, Garcia-Lazaro JA, Schnupp JWH (2009): Current source density profiles of stimulus-specific adaptation in rat auditory cortex. J. Neurophysiol. 102(3): 1483–1490.

35.

Antunes FM, Nelken I, Covey E, Malmierca MS (2010): Stimulus-specific adaptation in the auditory thalamus of the anesthetized rat. PLoS One. 5(11): e14071.

36.

Bäuerle P, von der Behrens W, Kössl M, Gaese BH (2011): Stimulus-specific adaptation in the gerbil primary auditory thalamus is the result of a fast frequency-specific habituation and is regulated by the corticofugal system. J. Neurosci. 31(26): 9708–9722.

37.

Taaseh N, Yaron A, Nelken I (2011): Stimulus-specific adaptation and deviance detection in the rat auditory cortex. PLoS One. 6(8): e23369.

38.

Duque D, Pérez-González D, Ayala YA, Palmer AR, Malmierca MS (2012): Topographic distribution, frequency, and intensity dependence of stimulus-specific adaptation in the inferior colliculus of the rat. J. Neurosci. 32(49): 17762–17774.

39.

Fishman YI, Steinschneider M (2012): Searching for the mismatch negativity in primary auditory cortex of the awake monkey: deviance detection or stimulus specific adaptation? J. Neurosci. 32(45): 15747–15758.

40.

Anderson LA, Malmierca MS (2013): The effect of auditory cortex deactivation on stimulus-specific adaptation in the inferior colliculus of the rat. Eur. J. Neurosci. 37(1): 52–62.

41.

Malmierca MS, Anderson LA, Antunes FM (2015): The cortical modulation of stimulusspecific adaptation in the auditory midbrain and thalamus: a potential neuronal correlate for predictive coding. Front. Syst. Neurosci. 9: 19.

23

42.

Camalier CR, Scarim K, Mishkin M, Averbeck BB (2019): A comparison of auditory oddball responses in dorsolateral prefrontal cortex, basolateral amygdala, and auditory cortex of macaque. J. Cogn. Neurosci. 31(7): 1054–1064.

43.

Javitt DC, Steinschneider M, Schroeder CE, Arezzo JC (1996): Role of cortical n-methyld-aspartate receptors in auditory sensory memory and mismatch negativity generation: implications for schizophrenia. Proc. Natl. Acad. Sci. USA. 93(21): 11962–11967.

44.

Umbricht D, Schmid L, Koller R, Vollenweider FX, Hell D, Javitt DC (2000): Ketamineinduced deficits in auditory and visual context-dependent processing in healthy volunteers: implications for models of cognitive deficits in schizophrenia. Arch. Gen. Psychiatry. 57(12): 1139–1147.

45.

Rosburg T, Kreitschmann-Andermahr I (2016): The effects of ketamine on the mismatch negativity (mmn) in humans - a meta-analysis. Clin. Neurophysiol. 127(2): 1387–1394.

46.

Ehrlichman RS, Maxwell CR, Majumdar S, Siegel SJ (2008): Deviance-elicited changes in event-related potentials are attenuated by ketamine in mice. J. Cogn. Neurosci. 20(8): 1403–1414.

47.

Sivarao DV, Chen P, Yang Y, Li Y-W, Pieschl R, Ahlijanian MK (2014): Nr2b antagonist cp-101,606 abolishes pitch-mediated deviance detection in awake rats. Front. Psychiatry. 5: 96.

48.

Harms L (2016): Mismatch responses and deviance detection in n-methyl-d-aspartate (nmda) receptor hypofunction and developmental models of schizophrenia. Biol. Psychol. 116: 75–81.

49.

Ahnaou A, Huysmans H, Biermans R, Manyakov NV, Drinkenburg WHIM (2017): Ketamine: differential neurophysiological dynamics in functional networks in the rat brain. Transl. Psychiatry. 7(9): e1237.

50.

Featherstone RE, Melnychenko O, Siegel SJ (2018): Mismatch negativity in preclinical models of schizophrenia. Schizophr. Res. 191: 35–42.

51.

Lee M, Balla A, Sershen H, Sehatpour P, Lakatos P, Javitt DC (2018): Rodent mismatch negativity/theta neuro-oscillatory response as a translational neurophysiological biomarker for n-methyl-d-aspartate receptor-based new treatment development in schizophrenia. Neuropsychopharmacology. 43(3): 571–582.

52.

Kantrowitz JT, Epstein ML, Lee M, Lehrfeld N, Nolan KA, Shope C, et al. (2018): Improvement in mismatch negativity generation during d-serine treatment in schizophrenia: correlation with symptoms. Schizophr. Res. 191: 70–79.

53.

Mitzdorf U (1985): Current source-density method and application in cat cerebral cortex: investigation of evoked potentials and eeg phenomena. Physiol. Rev. 65(1): 37–100.

24

54.

Steinschneider M, Reser D, Schroeder CE, Arezzo JC (1995): Tonotopic organization of responses reflecting stop consonant place of articulation in primary auditory cortex (a1) of the monkey. Brain Res. 674(1): 147–152.

55.

Lakatos P, Pincze Z, Fu K-MG, Javitt DC, Karmos G, Schroeder CE (2005): Timing of pure tone and noise-evoked responses in macaque auditory cortex. Neuroreport. 16(9): 933–937.

56.

Steinschneider M, Fishman YI, Arezzo JC (2008): Spectrotemporal analysis of evoked and induced electroencephalographic responses in primary auditory cortex (a1) of the awake monkey. Cereb. Cortex. 18(3): 610–625.

57.

Nelken I, Ulanovsky N (2007): Mismatch negativity and stimulus-specific adaptation in animal models. J. Psychophysiol. 21(3): 214–223.

58.

Schroeder CE, Steinschneider M, Javitt DC, Tenke CE, Givre SJ, Mehta AD, et al. (1995): Localization of erp generators and identification of underlying neural processes. Electroencephalogr. Clin. Neurophysiol. Suppl. 44: 55–75.

59.

Grimm S, Escera C, Nelken I (2016): Early indices of deviance detection in humans and animal models. Biol. Psychol. 116: 23–27.

60.

Jones EG (1985): The Thalamus, New York: Plenum Press,

61.

Javitt DC, Schroeder CE, Steinschneider M, Arezzo JC, Vaughan HG (1992): Demonstration of mismatch negativity in the monkey. Electroencephalogr. Clin. Neurophysiol. 83(1): 87–90.

62.

Komatsu M, Takaura K, Fujii N (2015): Mismatch negativity in common marmosets: whole-cortical recordings with multi-channel electrocorticograms. Sci. Rep. 5: 15006.

63.

Schwertner A, Zortea M, Torres FV, Caumo W (2018): Effects of subanesthetic ketamine administration on visual and auditory event-related potentials (erp) in humans: a systematic review. Front. Behav. Neurosci. 12: 70.

64.

Aitkin LM, Webster WR (1972): Medial geniculate body of the cat: organization and responses to tonal stimuli of neurons in ventral division. J. Neurophysiol. 35(3): 365–380.

65.

Wehr M, Zador AM (2003): Balanced inhibition underlies tuning and sharpens spike timing in auditory cortex. Nature. 426(6965): 442–446.

66.

Wu GK, Arbuckle R, Liu B-H, Tao HW, Zhang LI (2008): Lateral sharpening of cortical frequency tuning by approximately balanced inhibition. Neuron. 58(1): 132–143.

67.

O’Connell MN, Falchier A, McGinnis T, Schroeder CE, Lakatos P (2011): Dual mechanism of neuronal ensemble inhibition in primary auditory cortex. Neuron. 69(4): 805–817.

25

68.

Roger M, Arnault P (1989): Anatomical study of the connections of the primary auditory area in the rat. J. Comp. Neurol. 287(3): 339–356.

69.

Huang CL, Winer JA (2000): Auditory thalamocortical projections in the cat: laminar and areal patterns of input. J. Comp. Neurol. 427(2): 302–331.

70.

Yu X-J, Xu X-X, He S, He J (2009): Change detection by thalamic reticular neurons. Nat. Neurosci. 12(9): 1165–1170.

71.

Zikopoulos B, Barbas H (2007): Circuits for multisensory integration and attentional modulation through the prefrontal cortex and the thalamic reticular nucleus in primates. Rev Neurosci. 18(6):

72.

Teichert T, Gurnsey K (2019): Formation and decay of auditory short-term memory in the macaque monkey. J. Neurophysiol. 121(6): 2401–2415.

73.

Lakatos P, Chen C-M, O’Connell MN, Mills A, Schroeder CE (2007): Neuronal oscillations and multisensory interaction in primary auditory cortex. Neuron. 53(2): 279– 292.

74.

Lakatos P, Musacchia G, O’Connel MN, Falchier AY, Javitt DC, Schroeder CE (2013): The spectrotemporal filter mechanism of auditory selective attention. Neuron. 77(4): 750– 761.

75.

Sherman SM (2007): The thalamus is more than just a relay. Curr. Opin. Neurobiol. 17(4): 417–422.

76.

Lee CC, Sherman SM (2010): Drivers and modulators in the central auditory pathways. Front. Neurosci. 4: 79.

KEY RESOURCES TABLE Resource Type

Specific Reagent or Resource

Add additional rows as needed for each Include species and sex when applicable. resource type

Source or Reference Include name of manufacturer, company, repository, individual, or research lab. Include PMID or DOI for references; use “this paper” if new.

Antibody Bacterial or Viral Strain Biological Sample Cell Line Chemical Compound or Drug Commercial Assay Or Kit Deposited Data; Public Database Genetic Reagent Organism/Strain

Macaca Mulatta, 4 females 3 males

Lakatos Lab

Matlab 2018b

Mathworks, Natick, MA

Peptide, Recombinant Protein Recombinant DNA Sequence-Based Reagent Software; Algorithm Transfected Construct Other

Identifiers Include catalog numbers, stock numbers, database IDs or accession numbers, and/or RRIDs. RRIDs are highly encouraged; search for RRIDs at https://scicrunch.org/resources.

Additional Information Include any additional information or notes if necessary.

Lakatos et al.

Supplement

The Thalamocortical Circuit of Auditory Mismatch Negativity Supplemental Information Contents: Supplementary Methods and Materials Supplementary Discussion Supplementary References

SUPPLEMENTARY METHODS Subjects. All 7 macaque monkeys (5 female and 2 male) weighing 5-7 kg, who participated in the present study had been prepared surgically for chronic awake recordings. Prior to surgery, each animal was adapted to a custom fitted primate chair and to the recording chamber. All procedures were approved in advance by the Animal Care and Use Committee of the Nathan Kline Institute. Surgery. Preparation of subjects for awake intracortical recording was performed using aseptic techniques under general anesthesia. The tissue overlying the calvarium was resected and appropriate portions of the cranium were removed. The neocortex and overlying dura were left intact. To provide access to auditory cortical and thalamic regions, 2.5-centimeter dimeter plastic recording chambers (Crist Instrument Co. and Rogue Instruments) were positioned about 4 mm from the midline above both hemispheres. Together with socketed Plexiglas bars or custom headposts (to permit painless head restraint), they were secured to the skull with orthopedic screws and embedded in dental acrylic. A recovery time of six weeks was allowed before behavioral training and data collection began. Electrophysiology. During the experiments, animals were monitored with infrared and/or low light cameras as they sat in a primate chair in a dark, isolated, electrically shielded, soundattenuated chamber with head fixed in position. Neuroelectric activity was obtained using linear array multi-contact electrodes (23 contacts, 100 µm intercontact spacing, U- and V-probes, Plexon

1

Lakatos et al.

Supplement

Inc.). Neuroelectric signals were impedance matched with a pre-amplifier situated on the electrode (10x gain, bandpass dc-10 kHz), and after further amplification (500x) signals were recorded continuously with a 0.1 - 8000 Hz bandpass digitizer with a sampling rate of 20 kHz and precision of 16-bits using custom made software in Labview (National Instruments, Austin, TX) or with the SNR recording system (Alpha Omega, Alpharetta, GA), which had a sampling rate of 44 kHz. The signal was divided into field potential (0.1-300 Hz) and MUA (300-6000 Hz) range by zero phase shift digital filtering (2nd order Butterworth filter). MUA data was also rectified in order to improve the estimation of firing of the local neuronal ensemble. One-dimensional current source density (CSD) profiles were calculated from the local field potential profiles using a three-point formula for the calculation of the second spatial derivative of voltage (1). The advantage of CSD profiles is that they are not affected by volume conduction like the local field potentials, and they also provide a more direct index of the location, direction, and density of the net transmembrane current flow, at least in laminarly organized brain regions. Although MGB and pulvinar lack an apparent laminarly organized structure and therefore do not satisfy the basic assumptions underlying the application of one-dimensional CSD (2), sinks and sources indexing local net transmembrane current flow can still be observed in CSD response profiles. While the interpretation of these sinks and sources is not straightforward, their advantage remains that, as opposed to field potentials, they reflect strictly local synchronous neuronal events (3). At the beginning of each experimental session, after refining the electrode position in the neocortex, we established the best frequency (BF) of the A1 and thalamic recording site using a “suprathreshold” method (4–6), to make sure that we presented stimulus sequences that were either a frequency match or a mismatch (several octaves difference from BF, see below). The method entails presentation of a stimulus train consisting of 100 pseudorandom order occurrences of a broadband noise burst and pure tone stimuli with frequencies ranging from 353.5 Hz to 32 kHz in half octave steps (duration: 100 ms, r/f time: 5 ms, SOA = 624.5, loudness 50 dB). Auditory stimuli were produced using Labview (National Instruments, Austin, TX) at a 100 kHz sampling rate and presented via MF-1 free field speakers (TDT). During each experiment, cortical and thalamic responses to the presentation of pure tones, broadband noise bursts, clicks, strobe, and LED flashes were used to determine and refine electrode placement. Longer streams of pseudo-randomly presented pure tones and broadband noise bursts, as well as sound clouds were also used to determine the frequency tuning properties

2

Lakatos et al.

Supplement

of the recording sites. CSD and MUA response properties in all the above paradigms were used to functionally define cortical areas and thalamic nuclei offline. Identification of thalamic nuclei. To identify the different thalamic subdivisions described in the study, we used a combined anatomical/physiological/machine learning approach. The targeting of pulvinar vs. MGB nuclei is straightforward based on stereotaxic coordinates due to their relatively large anatomical separation (7). Reference stereotaxic coordinates in each subject are established during surgery by measuring the exact location and angle of the recording chambers. At the end of each animal’s experimental participation, functional assignment of the thalamic recording sites is confirmed histologically by the reconstruction of a subset of electrode tracks through postmortem histology (Nissl staining, parvalbumin and calbindin immunohistochemistry, e.g. Figure 1) following transcardial perfusion and whole brain sectioning. The data collected by verified electrode penetrations serve as ground truth (training dataset) for the supervised machine learning approach that is based on a variety of electrophysiological signal-based variables and is used to classify anatomically unverified electrode penetrations. For the machine learning based classification of thalamic recording sites, we currently extract 32 distinct CSD and MUA features partly from event related responses and partly from ongoing neuronal activity for each recording location (recording location corresponds to a single channel of the linear array multielectrode). These fall into the categories of onset latencies of responses to different auditory stimuli (broadband noise burst, click, and the tonotopy paradigm described above), response amplitudes, the latency of maximal response amplitude, sharpness of tuning, baseline MUA, and the spectral properties of ongoing CSD and MUA (i.e. the ratio of different frequency bands). Additionally, we also include mediolateral, anteroposterior and depth stereotaxic coordinates of the recordings. Using this 32-dimensional feature set, we train the cubic support vector machine algorithm using the MATLAB (Mathworks,Natick, MA) machine learning toolbox on the portion of the dataset with anatomically verified thalamic location. Based on a 5 folds cross-validation, the accuracy of our current algorithm is 94.2%. Finally, we assign the location of the rest of the data (e.g. recorded in animals that are still participating in experiments and therefore no anatomical data is available) using the trained model. Behavioral task and stimuli. While data recorded from 7 monkeys were utilized for the current experiments, none of them solely participated in the experiments described. Subjects were awake

3

Lakatos et al.

Supplement

during the recordings, but they were not required to respond to the auditory or visual stimuli presented. In all trial blocks analyzed, auditory and visual stimulus streams were presented simultaneously with differing stimulus onset asynchronies (SOAs) so that visual and auditory stimuli did not have a constant temporal relationship. The auditory stream consisted of pure tone beeps at 40 dB SPL (25 ms duration, 5 ms rise/fall time) with a constant SOA of 624.5 ms, equivalent to a 1.6 Hz repetition rate. The visual stream had a 1.8 Hz repetition rate. The frequency of the auditory standards was either parametrically varied across blocks in half octave steps between 0.3-32 kHz or was set to match the BF of the recording site or was set to a tone frequency at least 2 octaves different from the BF (non-BF streams). Frequency deviants (2-4 semitones different from the standard) occurred in the stream of standard tones every 3-9 seconds randomly. Maximally, 14 different frequency tone streams were presented per experiment, but after rejecting trial blocks with considerable movement artifacts (affecting more than 20% of deviant tone related responses), we were left with 1-13 trial blocks per recording site presented (mean = 3, std = 2.5). To be able to examine the effect of ketamine on deviant related thalamocortical responses and the surface MMN component, in a subset of experiments (12 A1 sites and 8 MGBd/m sites in 2 monkeys), we administered systemic subanesthetic doses of ketamine (1 mg/kg i.m.). After recording a control trial block, we injected the ketamine in the gluteus maximums of the subjects without removing them from the chair and recorded for 55 minutes at 10-minute intervals (6 postketamine trial blocks). Data analyses. Data were analyzed offline using native and custom-written functions in MATLAB (Mathworks, Natick, MA). CSD and MUA response profiles to a battery of auditory stimuli (see above) were used to determine the boundaries of cortical layers and thalamic nuclei (Fig. 1A). In order to examine stimulus specific adaptation, we analyzed how response amplitudes to the first 7 standard tones changed across trial blocks. In order to determine whether subsequent response amplitudes changed, we compared these to the amplitude of the first response of the stimulus train in the 5 -30 ms post-stimulus time interval (Fig 1). When comparing MUA and CSD responses to standard and deviant tones, we analyzed the same number of responses: responses to all deviant stimuli and responses to the standard that occurred just prior to each deviant (Figures 2-4). For time-frequency analyses (Figs. 3-4), continuous oscillatory amplitudes and phases were extracted by wavelet decomposition (Morlet wavelet, σ = 6). To characterize phase 4

Lakatos et al.

Supplement

distributions across trials, the wavelet transformed single trial data were normalized (unit vectors), the trials were averaged, and the length (modulus) of the resulting vector was computed (e.g. (8)). The value of the mean resultant length, also called inter-trial coherence (ITC), ranges from 0 to 1. Higher values indicated that the observations (oscillatory phase at a given time-point across trials) are clustered more closely around the mean than lower values (phase distribution is biased).

SUPPLEMENTARY DISCUSSION Our data clearly demonstrate that MMN generating mechanisms depend upon both subcortical and cortical processes. It appears that non-lemniscal MGB nuclei signal to the cortex that a deviant was detected by their layer 1 thalamocortical projections, possibly targeting calretinin/somatostatin inhibitory neurons which inhibit other inhibitory interneurons that target the distal apical dendrites of pyramidal neurons in Layer 1 (9, 10). This would result in a net disinhibitory effect, which could explain the deviant stimulus related L1 excitation described by our study. As recently suggested (11, 12), disinhibiting A1 pyramidal neurons across all of A1 would allow for their enhanced activation by deviant auditory stimulus related specific thalamocortical inputs. Since deviant related response onset latency in the pulvinar is similar to A1 regions not receiving specific standard or deviant stimulus related (granular layer) input, and pulvinar to A1 connectivity is sparse (13), we hypothesize that the role of the pulvinar is to promptly and widely disseminate deviant auditory stimulus related activity across higher order brain regions. We must emphasize here that our study focused on two thalamic structures (the MGB and pulvinar), and only the first stage of cortical auditory processing (A1). Therefore, the contribution of higher order cortical and subcortical brain regions to MMN generation is yet to be explored. A recent study by Camalier et al. (14) showed significant deviance-related activity in monkey DLPFC and amygdala that was ~50% the amplitude observed in auditory cortex. Similarly, insular contributions to MMN have been noted in humans (15), although the degree to which this activity contributes to scalp-recorded activity remains to be determined. We found that deviant-related activity can occur independently of SSA in A1 regions that do not receive lemniscal input related to the auditory frequency of either the standard or deviant stimulus, demonstrating critical differences between the two processes. As in the case of surface MMN recordings in humans (16–18), the spectral signature of MMN-like activity in NHP Layer

5

Lakatos et al.

Supplement

1 was confined mostly to the theta frequency range. Our results also indicate that oscillatory phase reset – a modulatory mechanism – plays a larger role in generating intracortical and surface MMN than an evoked type, driving mechanism, and it is strongly affected by ketamine. Because we used systemic ketamine injection, the locus of ketamine effect cannot be determined.

However, we have previously demonstrated that local infusion of NMDAR

antagonists into auditory cortex significantly disrupts MMN generation (as in the present study and (19, 20)), suggesting that blockade of cortical NMDAR is sufficient to induce schizophrenialike deficits, although additional dysfunction within subcortical regions such as inferior colliculus may also contribute (21). In general, the current flow that gives rise to surface-recordable activity arises from spatially aligned dendrites of pyramidal neurons. Ketamine-induced MMN inhibition may therefore reflect reduced NMDAR dependent current flow through open, unblocked NMDAR channels. In addition, it has recently been demonstrated that local inactivation of somatostatin, but not parvalbumin type GABAergic interneurons inhibits visual MMN generation in rodents (10). Thus, additional effects on somatostatin interneurons may also be critical. Future studies with intracranial and cell-specific manipulations are needed to further probe underlying circuits (22).

6

Lakatos et al.

Supplement

SUPPLEMENTARY REFERENCES 1.

Mitzdorf U (1985): Current source-density method and application in cat cerebral cortex: investigation of evoked potentials and eeg phenomena. Physiol. Rev. 65(1): 37–100.

2.

Nicholson C, Freeman JA (1975): Theory of current source-density analysis and determination of conductivity tensor for anuran cerebellum. J. Neurophysiol. 38(2): 356– 368.

3.

Barczak A, O’Connell MN, McGinnis T, Ross D, Mowery T, Falchier A, et al. (2018): Topdown, contextual entrainment of neuronal oscillations in the auditory thalamocortical circuit. Proc. Natl. Acad. Sci. USA. 115(32): E7605–E7614.

4.

Steinschneider M, Fishman YI, Arezzo JC (2008): Spectrotemporal analysis of evoked and induced electroencephalographic responses in primary auditory cortex (a1) of the awake monkey. Cereb. Cortex. 18(3): 610–625.

5.

Steinschneider M, Reser D, Schroeder CE, Arezzo JC (1995): Tonotopic organization of responses reflecting stop consonant place of articulation in primary auditory cortex (a1) of the monkey. Brain Res. 674(1): 147–152.

6.

Lakatos P, Pincze Z, Fu K-MG, Javitt DC, Karmos G, Schroeder CE (2005): Timing of pure tone and noise-evoked responses in macaque auditory cortex. Neuroreport. 16(9): 933–937.

7.

Saleem KS, Logothetis NK (2012): A Combined Mri And Histology Atlas Of The Rhesus Monkey Brain In Stereotaxic Coordinates,, 2, ed. Academic Press,, p. 402

8.

O’Connell MN, Barczak A, Ross D, McGinnis T, Schroeder CE, Lakatos P (2015): Multiscale entrainment of coupled neuronal oscillations in primary auditory cortex. Front. Hum. Neurosci. 9: 655.

9.

Cruikshank SJ, Ahmed OJ, Stevens TR, Patrick SL, Gonzalez AN, Elmaleh M, et al. (2012): Thalamic control of layer 1 circuits in prefrontal cortex. J. Neurosci. 32(49): 17813–17823.

10.

Hamm JP, Yuste R (2016): Somatostatin interneurons control a key component of mismatch negativity in mouse visual cortex. Cell Rep. 16(3): 597–604.

11.

Larkum M (2013): A cellular mechanism for cortical associations: an organizing principle for the cerebral cortex. Trends Neurosci. 36(3): 141–151.

12.

Abs E, Poorthuis RB, Apelblat D, Muhammad K, Pardi MB, Enke L, et al. (2018): Learningrelated plasticity in dendrite-targeting layer 1 interneurons. Neuron. 100(3): 684–699.e6.

13.

Scott BH, Saleem KS, Kikuchi Y, Fukushima M, Mishkin M, Saunders RC (2017): Thalamic connections of the core auditory cortex and rostral supratemporal plane in the macaque monkey. J. Comp. Neurol. 525(16): 3488–3513.

7

Lakatos et al.

Supplement

14.

Camalier CR, Scarim K, Mishkin M, Averbeck BB (2019): A comparison of auditory oddball responses in dorsolateral prefrontal cortex, basolateral amygdala, and auditory cortex of macaque. J. Cogn. Neurosci. 31(7): 1054–1064.

15.

Kantrowitz JT, Hoptman MJ, Leitman DI, Moreno-Ortega M, Lehrfeld JM, Dias E, et al. (2015): Neural substrates of auditory emotion recognition deficits in schizophrenia. J. Neurosci. 35(44): 14909–14921.

16.

Hochberger WC, Joshi YB, Zhang W, Thomas ML, Consortium of Genomics in Schizophrenia (COGS) investigators, Braff DL, et al. (2018): Decomposing the constituent oscillatory dynamics underlying mismatch negativity generation in schizophrenia: distinct relationships to clinical and cognitive functioning. Int. J. Psychophysiol.

17.

Javitt DC, Lee M, Kantrowitz JT, Martinez A (2018): Mismatch negativity as a biomarker of theta band oscillatory dysfunction in schizophrenia. Schizophr. Res. 191: 51–60.

18.

Javitt DC, Shelley A-M, Ritter W (2000): Associated deficits in mismatch negativity generation and tone matching in schizophrenia. Clin. Neurophysiol. 111(10): 1733–1737.

19.

Javitt DC, Steinschneider M, Schroeder CE, Vaughan HG, Arezzo JC (1994): Detection of stimulus deviance within primate primary auditory cortex: intracortical mechanisms of mismatch negativity (mmn) generation. Brain Res. 667(2): 192–200.

20.

Javitt DC, Steinschneider M, Schroeder CE, Arezzo JC (1996): Role of cortical n-methyl-daspartate receptors in auditory sensory memory and mismatch negativity generation: implications for schizophrenia. Proc. Natl. Acad. Sci. USA. 93(21): 11962–11967.

21.

Gaebler AJ, Zweerings J, Koten JW, König AA, Turetsky BI, Zvyagintsev M, et al. (2019): Impaired subcortical detection of auditory changes in schizophrenia but not in major depression. Schizophr. Bull.

22.

Javitt DC, Freedman R (2015): Sensory processing dysfunction in the personal experience and neuronal machinery of schizophrenia. Am. J. Psychiatry. 172(1): 17–31.

8