Stimulus-specific regulation of visual oddball differentiation in posterior parietal cortex — TEI-tagged view

ZhoupersName, HuangpersName, YupersName, NegahbanipersName, StittpersName, AlexanderpersName, HammpersName, KatopersName (2020date). “Stimulus-specific regulation of visual oddball differentiation in posterior parietal cortex.” Scientific Reportstitle·j.

Object of study — how single neurons in a higher-order visual area distinguish a frequently presented stimulus from a rareterm one, and whether that depends on which stimulus the local population prefers

Research question — whether the mismatch response is driven by adaptationterm to the standardterm or by an enhanced response to the deviantterm, and whether it depends on stimulus tuningterm

Contribution — the mismatch response appears only for the population-preferred stimulus and is carried almost entirely by adaptationterm to the standardterm rather than by surprise at the deviantterm

Full text — tagged

The reference list is not part of the text we hold for this paper. Our copy was extracted from the publisher's page and the bibliography did not survive that extraction, so the article below ends before its references. Nothing has been removed by us.

Stimulus-specific regulation of visual oddballterm·mth differentiation in posterior parietal cortex Zhe Charles ZhoupersName Zhe Charles ZhoupersName 1 Department of Psychiatry, University of North Carolina at Chapel Hill, 116 Manning Drive, 6018A, Chapel Hill, NC 27599 USA 5 Neuroscience Center, University of North Carolina at Chapel Hill, Chapel Hill, NC 27599 USA 6 Neurobiology Curriculum, University of North Carolina at Chapel Hill, Chapel Hill, NC 27599 USA 7 Carolina Center for Neurostimulation, University of North Carolina, Chapel Hill, NC 27599 USA 1, 5, 6, 7 , Wei Angel HuangpersName Wei Angel HuangpersName 1 Department of Psychiatry, University of North Carolina at Chapel Hill, 116 Manning Drive, 6018A, Chapel Hill, NC 27599 USA 5 Neuroscience Center, University of North Carolina at Chapel Hill, Chapel Hill, NC 27599 USA 6 Neurobiology Curriculum, University of North Carolina at Chapel Hill, Chapel Hill, NC 27599 USA 7 Carolina Center for Neurostimulation, University of North Carolina, Chapel Hill, NC 27599 USA 1, 5, 6, 7 , Yiyi Yu Yiyi Yu 9 Department of Biomedical Sciences, University of California at Santa Barbara, Los Angeles, CA 90048 USA 9 , Ehsan NegahbanipersName Ehsan NegahbanipersName 1 Department of Psychiatry, University of North Carolina at Chapel Hill, 116 Manning Drive, 6018A, Chapel Hill, NC 27599 USA 7 Carolina Center for Neurostimulation, University of North Carolina, Chapel Hill, NC 27599 USA 1, 7 , Iain M StittpersName Iain M StittpersName 1 Department of Psychiatry, University of North Carolina at Chapel Hill, 116 Manning Drive, 6018A, Chapel Hill, NC 27599 USA 7 Carolina Center for Neurostimulation, University of North Carolina, Chapel Hill, NC 27599 USA 1, 7 , Morgan L AlexanderpersName Morgan L AlexanderpersName 1 Department of Psychiatry, University of North Carolina at Chapel Hill, 116 Manning Drive, 6018A, Chapel Hill, NC 27599 USA 7 Carolina Center for Neurostimulation, University of North Carolina, Chapel Hill, NC 27599 USA 1, 7 , Jordan P HammpersName Jordan P HammpersName 8 Neuroscience Institute, Georgia State University, Atlanta, GA 30302 USA 8 , Hiroyuki K KatopersName Hiroyuki K KatopersName 1 Department of Psychiatry, University of North Carolina at Chapel Hill, 116 Manning Drive, 6018A, Chapel Hill, NC 27599 USA 5 Neuroscience Center, University of North Carolina at Chapel Hill, Chapel Hill, NC 27599 USA 6 Neurobiology Curriculum, University of North Carolina at Chapel Hill, Chapel Hill, NC 27599 USA 10 Carolina Institute for Developmental Disabilities, University of North Carolina at Chapel Hill, Chapel Hill, NC 27599 USA 1, 5, 6, 10 , Flavio Fröhlich Flavio Fröhlich 1 Department of Psychiatry, University of North Carolina at Chapel Hill, 116 Manning Drive, 6018A, Chapel Hill, NC 27599 USA 2 Department of Cell Biology and Physiology, University of North Carolina at Chapel Hill, Chapel Hill, NC 27599 USA 3 Department of Biomedical Engineering, University of North Carolina at Chapel Hill, Chapel Hill, NC 27599 USA 4 Department of Neurology, University of North Carolina at Chapel Hill, Chapel Hill, NC 27599 USA 5 Neuroscience Center, University of North Carolina at Chapel Hill, Chapel Hill, NC 27599 USA 6 Neurobiology Curriculum, University of North Carolina at Chapel Hill, Chapel Hill, NC 27599 USA 7 Carolina Center for Neurostimulation, University of North Carolina, Chapel Hill, NC 27599 USA 1, 2, 3, 4, 5, 6, 7, ✉ 1 Department of Psychiatry, University of North Carolina at Chapel Hill, 116 Manning Drive, 6018A, Chapel Hill, NC 27599 USA 2 Department of Cell Biology and Physiology, University of North Carolina at Chapel Hill, Chapel Hill, NC 27599 USA 3 Department of Biomedical Engineering, University of North Carolina at Chapel Hill, Chapel Hill, NC 27599 USA 4 Department of Neurology, University of North Carolina at Chapel Hill, Chapel Hill, NC 27599 USA 5 Neuroscience Center, University of North Carolina at Chapel Hill, Chapel Hill, NC 27599 USA 6 Neurobiology Curriculum, University of North Carolina at Chapel Hill, Chapel Hill, NC 27599 USA 7 Carolina Center for Neurostimulation, University of North Carolina, Chapel Hill, NC 27599 USA 8 Neuroscience Institute, Georgia State University, Atlanta, GA 30302 USA 9 Department of Biomedical Sciences, University of California at Santa Barbara, Los Angeles, CA 90048 USA 10 Carolina Institute for Developmental Disabilities, University of North Carolina at Chapel Hill, Chapel Hill, NC 27599 USA ✉ Corresponding author.

Received 2020 Mar 14; Accepted 2020 Jul 22; Collection date 2020.

© The Author(s) 2020 Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptationterm, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons license, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/ .

PMCID: PMC7435179 PMID: 32811878 Abstract The frequency at which a stimulus is presented determines how it is interpreted. For example, a repeated image may be of less interest than an image that violates the prior sequence. This process involves integration of sensory information and internal representations of stimulus history, functions carried out in higher-order sensory areas such as the posterior parietal cortex (PPCterm). Thus far, there are few detailed reports investigating the single-neuron mechanisms for processing of stimulus presentation frequency in PPCterm. To address this gap in knowledge, we recorded PPCterm activity using 2-photon calcium imaging and electrophysiologyterm·mth during a visual oddball paradigmterm·def. Calcium imaging results reveal differentiation at the level of single neurons for frequent versus rareterm conditions which varied depending on whether the stimulus was preferredterm or non-preferred by the recorded neural population. Such differentiation of oddballterm·mth conditions was mediated primarily by stimulus-independent adaptationterm in the frequent condition.

Subject terms: Neuroscience, Sensory processing Introduction Behavioral flexibility requires the brain to generate distinct patterns of neural responses when a stimulus is observed frequently versus rarely. For example, a wild deer that encounters a car on a daily basis may perceive rareterm, intermittent encounters differently. The latter case is much more surprising since the stimulus (driving car) possesses substantially different features from the usual experienced environment. Accordingly, there must be circuitry in the brain that is able to differentiate such scenarios.

The oddball paradigmterm·def is a classic sensory stimulus assay used to evaluate neural differentiation of stimulus presentation frequency 1 , 2 . In this paradigm, the neural response to a rareterm, deviantterm stimulus is compared to the response to a frequently-presented standardterm stimulus; differentiation of the two conditions is quantified as the difference in responses between the standardterm and deviantterm (called mismatch negativity). Specifically, the mismatch negativity (MMN)term·def is a neurophysiological difference waveform peaking in the 150–250 ms window after stimulus onset. The MMNterm refers to the difference in evoked response to the standardterm and deviantterm stimulus. Importantly, the MMNterm is distinct from the N100 (occurring between 50 and 150 ms post stimulus onset) and the P300 (occurring around 300–400 ms) components with respect to function and timing relative to stimulus onset 3 , 4 . The MMNterm has been studied comprehensively in humans using EEG, MEG, and fMRI 5 – 10 , and to a lesser extent, in animal models with local field potential (LFPterm) and/or single units 11 – 16 . Yet, few studies have extensively probed how individual neurons exhibit responses to oddballterm·mth conditions as a function of stimulus properties. Stimulus preference and tuningterm in single neurons may influence MMNterm responses, highlighting the need for neural recordings at higher spatial granularity using methods such as 2-photon calcium imaging.

It has been postulated that the MMNterm is a precursor to sensory discrimination 17 , 18 . Within this framework, sensory areas such as visual or parietal cortex may show differing levels of activity to the surprising deviantterm stimulus versus the frequently presented standardterm stimulus, which both hold behavioral relevance to the animal. However, whether the MMNterm is dominated by adaptationterm to the standardterm condition or an enhanced response to the deviantterm condition remains a matter of debate 17 , 19 . Further, heterogeneity in stimulus tuningterm across neurons in sensory cortices may affect oddballterm·mth responses where neurons that show stronger baseline responses to a particular stimulus (e.g., motion in one directionterm versus the other) exhibit larger magnitude MMNterm compared to a stimulus with weaker baseline responses. Accordingly, neurons encoding different stimulus features effectively compete to relay behaviorally-relevant information to downstream areas.

Posterior parietal cortex (PPCterm) is well positioned in the visual circuit to transform basic sensory tuningterm to signals reflecting the frequency of stimulus presentation due to its functional involvement in stimulus motion processing and history-dependent evidence accumulation 20 – 22 . Importantly, PPCterm specializes in differentiating stimulus motion directionsterm·var: upstream visual brain regions from PPCterm encode orientationterm tuningterm, whereas sensory responses in PPCterm show stronger responses to one directionterm over the opposite. Further, single neurons in PPCterm can exhibit unique tuningterm to different directionsterm 22 , 23 . Given this directionterm tuningterm hallmark of PPCterm function, MMNterm responses may differ as a function of a neuron’s directionterm tuningterm in this brain area.

To understand how PPCterm neurons modulate their activity in response to stimuli presented in frequent and rareterm conditions, we recorded extracellular electrophysiologyterm·mth and performed 2-photon calcium imaging in PPCterm of head-fixed ferrets during a drifting grating visual oddball paradigmterm·def. Our results indicate that PPCterm LFPterm and single neuron calcium imaging activity exhibit stimulus-specific modulation across oddballterm·mth conditions. Spectral decomposition of the LFPterm revealed MMNterm in the gammaterm frequency band for the preferredterm stimulus only. We leveraged the strength of 2-photon calcium imaging by comparing oddballterm·mth activity for each spatially-resolved neuron and found that mismatch responses (MMRterm, equivalent to MMNterm but for calcium imaging data) scaled in magnitude with neuronal tuningterm to the population-preferred stimulus. We further dissected the components of the MMRterm and discovered that it was dominated primarily by stimulus-specific adaptationterm·def.

Materials and methods Subjects A total of six adult (16–19 weeks old) female ferrets ( Mustela putorius furo , weighing 0.7–1 kg, group housed in a 12 h light/12 h dark cycle; spayed, Marshall BioResources, North Rose, NY) were included in this study. Three animals were included in the electrophysiological experiment (across these animals, we acquired 15 sessions) and four were included in the 2-photon calcium imaging experiment (across these animals, we acquired 21 sessions). All animal procedures were performed in compliance with the National Institutes of Health guide for the care and use of laboratory animals (NIH Publications 8th edition, 2011), and the United States Department of Agriculture (USDA Animal Welfare).

Electrophysiologyterm·mth surgery Each animal was initially anesthetized with a ketamine/xylazine cocktail (30 mg/kg of ketamine, 1–2 mg/kg of xylazine, intramuscular) and was checked for toe-pinch response for a stable, deep plane of anesthesia before continuing with intubation with an endotracheal tube for subsequent artificial ventilation and anesthetic maintenance (0.5–2% isoflurane in 100% oxygen). The animal was then positioned into a stereotaxic frame to allow for precise measurement of target brain regions and to provide stability of the head throughout the surgical procedure. All following surgical procedures were performed under asceptic conditions. After applying alcohol and betadine swabs to the scalp, the skin and muscle overlying the skull were cut and resected. A custom-machined headpost was secured to the area of the skull above the right parietal cortex and dental cement (C&B metabond, Parkell, Edgewood, NY) was applied to the base of the headpost. Craniotomy for the left parietal cortex was drilled (13.5 mm anterior to the caudal crest landmark, 3–4 mm lateral from midline) and dura mater and pia mater were resected. Sixteen channel electrode arrays (200 μm distance between each electrode; local reference electrode 500 μm shorter than recording electrodes; Innovative Neurophysiology, Durham, NC) were then lowered into cortical tissue at the depth of around 600 μm. Dental acrylic (Lang Dental, Wheeling, IL) was applied to secure electrodes and seal the craniotomy. Skin, connective tissue, and muscle were then sutured together; triple antibiotic ointment was applied to the wound margin and animals were administered pain relief medication (meloxicam, 0.2 mg/kg). Post-operative care included pain medication (meloxicam, 0.2 mg/kg) and infection controlterm (12.5–13 mg/kg, clavamox, Zoetis, Parsippany, NJ) in addition to headcap cleaning.

Calcium imaging surgery Animals in the calcium imaging group (n = 4) underwent two surgeries: virus injection and cranial window implantation. Surgery procedures for the virus injection were similar to that of the electrophysiologyterm·mth group with the following differences: a smaller craniotomy above PPCterm was drilled (~ 2–3 mm in diameter). After the dura was resected, a mix of 800 nL of AAV9.CaMKII.GCaMP6f.WPRE.SV40 (viral titer = 7.55e12, UNC Viral Vector Core) and 200 nL of 1% fast green in saline was injected into cortical tissue at a depth of about 200–400 μm below cortex using a glass pipette injector system (Nanoject, Drummond Scientific, Broomall, PA). The virus/fast green solution was injected in 31 bouts of 32.2 nL at 46 nL/s with about 15 s of wait time between each bout. The tip of the glass pipette was sharp enough to penetrate through the intact pia mater with minimal damage. The craniotomy was then sealed using kwik-kast (World Precision Instruments, Sarasota, FL); then muscle, tissue, and skin were sutured back together. Cranial window surgeries were closely modelled based on a published methods paper 24 .

After around 3–5 weeks of virus expression time, the same animals underwent the cranial window surgery. The goal of the cranial window surgery was to make a larger craniotomy around the virus expression site and implant a custom coverslip holder (a machined metal ring-like structure that holds and secures a coverslip) and a glass coverslip above cortical tissue to allow for optical access. After animals were prepared for asceptic surgery and tissue overlying the skull was resected, an 8 mm circular area around the virus injection craniotomy was shaved down; this ultimately allows the coverslip structure to be as close to the brain as possible. The coverslip holder was then secured to the skull using C&B metabond and the exposed skull in the middle of the ring was then drilled out. Dura was then resected and a coverslip assembly (an 8 mm, 200 μm thick circular coverslip optical glued to a 5 mm, 1.4 mm thick cylindric glass piece; Swiftglass, Elmira, NY) was lowered into the coverslip holder. Importantly, the coverslip pressed firmly on the brain which prevents dura regrowth under the coverslip and stabilizes the brain during imaging. A retaining ring (McMaster-Carr, Atlanta, GA) was secured above the glass coverslip and superglue was applied to seal the structure. Recording sessions were made about 8–12 weeks post virus injection surgery. It is possible that such a period of time between virus injection and data acquisition may result in overexpression of GCaMP6f; however upon inspection of mean fluorescence images (Supplementary Fig. 1 ), we found at most 2–3 cells out of each animal’s FOV to exhibit nuclear filling.

Oddballterm·mth stimulation paradigm Prior to surgery, animals were habituated to a custom, 3d-printed fixation setup that included a body holding tube and a head-fixation post (Thorlabs posts, Thorlabs, Newton, NJ; and custom-machined head-post clamp). After habituation and electrode/window implantation surgery, visual oddballterm·mth stimulation and recordings commenced. Visual stimulation was displayed via presentation software (Neurobehavioral Systems, Berkeley, CA) on a computer monitor (120 frame rate) positioned 40 cm away from the animal. The oddballterm·mth stimulation paradigm was designed as follows: stimuli consisted of grayscale full-field sine wave drifting gratings with a spatial frequencyterm·var of 0.2 cycles per degree and drifting at 4 Hz. On the first day of neurophysiological recordings, animals were administered a session consisting of 10 drifting gratings of different motion directionsterm·var (0°, 36°, 72°, 108°, 144°, 180°, 216°, 252°, 288°, 324°) presented at equiprobability to assess which stimuli elicited the most activity from the population. Typically the stimulus evoking the largest amplitude response during stimulus presentation (0–400 ms) and opposing directionterm stimulus were chosen for the following sessions. The following oddball paradigmterm·def sessions consisted of three randomized blocks of 300 trials where each trial consisted of stimulus presentation for 400 ms and an equi-luminance gray screen inter-trial interval (ITI) following for 1,000–1,500 ms. Trial order was randomized for each block. In one block (oddballterm·mth A), the two best responsive stimuli from the first session were chosen to be presented in either the standardterm (90% of trials) or deviantterm (10% of trials) condition (ie. stimulus A was the deviantterm and B was the standardterm). In another block (oddballterm·mth B), the same two stimuli were used but the conditions were switched (ie. stimulus B was the deviantterm and A was the standardterm). The reason for presenting both of these blocks is to be able to compare the same stimulus presented in different conditions, thus removing the confounding factor of stimulus-preference when examining neural activity. A third block’s (many-standards control blockterm·var) stimuli consisted of 10 equiprobable drifting gratings of different motion directionsterm·var (same structure and stimuli as the initial tuningterm curve session); stimuli presented in this third condition served as controlsterm that do not elicit adaptationterm effects of the standardterm condition or surprise effects like the deviantterm condition. A series of random salt-and-pepper images (white noise checkerboard—10 × 6 squares, 33.3 ms presentation with 1,000–1,500 ms ITI gray screen) were presented in between the oddball paradigmterm·def blocks to reset the visual system and minimize adaptationterm (method adapted from a previous publication 15 ).

Electrophysiologyterm·mth sessions Animals were first placed into a custom 3d-printed tube and head-fixed using a custom-machined head-post clamp. Broadband signals (1 Hz hardware highpass cutoff) from PPCterm were acquired at a 30 kHz sampling rate and digitized using an electrophysiologyterm·mth data acquisition system (Intan Technologies, Los Angeles, CA). The electrophysiologyterm·mth system also recorded TTL pulses sent from a separate computer running the visual stimulation via a data acquisition system (DAQ, Texas Instruments, Dallas, TX). All recording sessions were performed in a dark room save the computer monitor that displayed visual stimuli.

Two-photon imaging Recording of optically-resolved population GCaMPterm activity was performed using a resonant scanner, 2-photon microscope (Neurolabware, Los Angeles, CA), 16 × water-immersion objective (Nikon, 0.8 NA, 3 mm working distance), and the Scanbox acquisition software (Scanbox, Los Angeles, CA). Two-photon excitation of GCaMPterm fluorophores was delivered through the microscope using a Mai-tai Tai Sapphire laser (Spectra Physics, Santa Clara, CA) at 940 nm, and light emission was collected using a GaAsP photo-multiplier tube. Data were acquired at the frame rate of 15.49 fps. At the beginning of each session, the objective was lowered to a focal plane of 150–200 μm below the cortical surface, around layer II/III of the ferret cortex 25 . Data were typically acquired at 1.4 × digital zoom, resulting in a field of view length and width of about 800 um.

Electrophysiologyterm·mth analysis All analyses were performed in Matlab (Mathworks, Natick, MA). All broadband electrophysiological data were preprocessed in three different ways for event-related potential (ERPterm), spectral decomposition, and spikingterm analysis. For ERPterm analysis, broadband data were first low-pass filtered at 30 Hz using a 4th order, phase-preserving butterworth filter then downsampled to 300 Hz. For spectral analysis, broadband data were first low-pass filtered at 300 Hz using a 4th order, phase-preserving butterworth filter then downsampled to 300 Hz. For spikingterm analysis, broadband data were first band-pass filtered between 300 and 5,000 Hz using a 4th order butterworth filter. Spike events were detected in each channel whenever the signal crossed minus 4 standard deviationsterm·stat.

ERPsterm were calculated by extracting data in a window around each trial onset, then baseline corrected by subtracting the mean activity from − 100 to 0 ms relative to stimulus onset. Note that all times reported are relative to stimulus onset. Data were first averaged across trials for each oddballterm·mth condition and stimulus, then averaged across recording channels and sessions. Statistical comparisons and standard error of the mean (SEM)term·stat were calculated across sessions. Condition contrasts (MMNterm: deviant-standard, DDterm: deviant-control) were calculated at the level of channels after averaging over trials.

For spectral analysis, we wanted to examine changes in ongoing oscillations that are not phase-locked to stimulus presentation (induced oscillations). Accordingly, the average ERPterm across trials was calculated and subtracted from each trial before subsequent analysis. The analytic signal for each frequency was then calculated by convolving a family of Morlet waveletsterm·stat with the LFPterm from 1 to 50 Hz in 0.5 Hz increments. Power was then calculated as the absolute value of the analytic signal squared. Data for each trial were then extracted and averaged for each condition and stimulus. Finally, data were averaged across recording channels and sessions. Stats and standard error of the mean (SEM)term·stat were calculated across sessions.

Calcium imaging analysis Calcium imaging video data were first preprocessed using the Suite2p toolbox 26 . Using this toolbox, images were aligned using non-rigid motion correction and subject to singular value decomposition (SVD)term·stat across time to reduce dimensionality and reduce subsequent computation time. Pixels in the image were then clustered into regions-of-interest (ROIsterm) based on optimization of their spatial footprint and activity time-series using a model loosely based on expectation–maximization (EM). Using these ROIsterm, fluorescence activity was calculated by first taking the average time-series across pixels within the ROIterm; then neuropilterm background activity was estimated by defining a ring area around the ROIterm (inner diameter = 3 pixels away from ROIterm border, outer diameter = 5 pixels away), calculating principal component analysis (PCA)term·stat across pixels in the ring, and taking the first principal component’s (PC) activity; finally neuropil-corrected time-courses were calculated by subtracting the neuropilterm 1st PC activity from that of the ROIterm. Change in fluorescence was defined as Δ F / F t = ( F t - F avg ) / F avg where F t is the fluorescence at a given time t and F avg is the average fluorescence across the whole session.

ROIsterm that exhibited stimulus-evoked activity (300–1,500 ms) 1.5 standard deviationsterm·stat above baseline (− 200 to 0 ms) mean activity for at least 400 consecutive milliseconds for at least 25% of trials, and 1.5 standard deviationsterm·stat above baseline mean activity for at least 400 consecutive milliseconds in the trial average were identified as visually responsive and were included for further analysis (n = 73). Activity for each trial was then extracted and averaged for each condition and stimulus. All of the calcium imaging preprocessing steps thus far are summarized in Fig. 3 D. Time-courses were baseline corrected by subtracting the mean activity from − 200 to 0 s relative to stimulus onset. Data were averaged across recording ROIsterm and sessions. Statistics and standard error of the mean (SEM)term·stat were calculated across ROIsterm. Due to the slow decay kinetics of the GCaMP6f fluorescence indicator, it is difficult to interpret when certain processes are occurring during periods of neural activity. Accordingly, we chose the period of time that calcium responses largely decayed back to baseline (from 300 to 1,500 ms relative to stimulus onset) when performing statistics or averaging across the epoch of time related to stimulus response. The comparison between standardterm and deviantterm conditions for the calcium imaging data was called mismatch response (MMR)term·def due to the GCaMPterm protein exhibiting primarily positive increases in activity and its slow decay kinetics. Condition contrasts (MMRterm: deviant-standard, DDterm: deviant-control) were calculated at the level of ROIsterm after averaging over trials.

Figure 3.

2-photon calcium imaging reveals stimulus-specific MMRterm in PPCterm. ( A ) Schematic of cranial window implant and calcium imaging setup. ( B ) Sample coronal section of PPCterm expression of GCaMP6f. Bold labels represent anatomical region names. Italicized labels represent functional region names. Red scale bar represents 5 mm distance. Inset: close up of recording area with cells labeled with GCaMP6f. Abbreviations: lateral gyrus (LG), posterior parietal cortex (PPCterm), cingulate gyrus (CG), retrosplenial cortex (RSC), suprasylvian gyrus (SSG), middle ectosylvian gyrus (MEG), hippocampus (Hip). Labels based on published ferret brain atlas 66 . ( C ) Left: Sample max fluorescence projection of a calcium imaging recording. Black contours represent regions of interest (ROIsterm) that were identified using a calcium imaging cell detection algorithm (Suite2p). Insets are close-ups of sample ROIsterm. Colored contours correspond to the sample fluorescence activity traces shown on the right. Signal extraction from ROIsterm was performed using a custom-designed algorithm with neuropilterm subtraction and dF/F calculation. ( D ) Calcium imaging preprocessing pipeline. Raw calcium imaging frames were first motion corrected then underwent ROIterm segmentation using the Suite2p toolbox 67 . ROIterm activity time-series were calculated by averaging across ROIterm pixels and subtracting the corresponding background (BG) activity defined by pixels in a ring around the ROIterm. Normalized activity was calculated as change in fluorescence over mean fluorescence across the session. To ensure all ROIsterm defined plausible cells, we manually curated the data based on ROIterm contours and time-series. Finally visually-responsive ROIsterm were chosen for subsequent analysis based on a stringent threshold for stimulus-evoked responses. ( E ) To assess directionterm tuningterm at the level of single neurons, head-fixed ferrets were presented a series of drifting gratings during 2-photon calcium imaging sessions. ( F ) Trial- and ROI-averaged event-related calcium responses for each drifting grating directionterm (in color) for a representative animal. Black bar represents the duration of stimulus presentation. ( G ) Polar histogram of normalized time-averaged (300–1,500 ms) calcium responses as a function of drifting grating directionterm. Note the two opposing directionsterm that elicit the largest responses: the preferredterm and opponentterm stimulus.

A concern arises when comparing activity between conditions for a certain stimulus: observed differences could arise and be confounded by block-to-block differences in baseline fluorescence. To account for this concern, we baseline corrected all time-courses on a trial level. Further, when calculating the condition contrasts, we performed a trial condition shuffle controlterm·stat. This first involved computing trial-averaged activity with trial conditions shuffled within each block (effectively preserving block-to-block differences in activity, while eliminating condition-dependent differences). The condition contrast was then computed (MMRterm or DDterm). Finally, trial-averaged shuffleterm·mth data were subtracted from trial condition-intact data on a ROI-by-ROI basis.

To quantify an ROIterm’s stimulus preference, a modulation indexterm was calculated: S pref - S Opp S pref + S Opp where S Pref and S Opp corresponded to the trial-averaged response during the 300–1,500 ms epoch to the preferred stimulus and opponent stimulusterm·def, respectively. This modulation indexterm was calculated for both the MMRterm and controlterm activity for each ROIterm.

All figures were prepared using Adobe Illustrator CS6 (version 16.0.3; San Jose, CA).

Ethical approval All animal procedures were performed in compliance with the National Institutes of Health guide for the care and use of laboratory animals (NIH Publications 8th edition, 2011), and approved by the Institutional Animal Care and Use Committee of the University of North Carolina at Chapel Hill and the United States Department of Agriculture (USDA Animal Welfare).

Results To dissect how processing of oddballterm·mth conditions occurs as a function of stimulus preference, we first delineated stimulus tuningterm by presenting a series of randomized drifting gratings to head-fixed ferrets during extracellular electrophysiologyterm·mth (Fig. 1 A, n = 3 animals) recording sessions. We found that ferret posterior parietal cortex (PPCterm) firing activity and local field potentials (LFPterm) primarily showed strong responses to two opposing directionsterm, similar to reports in other animal species (Fig. 1 B,D) 20 , 22 . The stimulus directionterm that elicited the largest normalized amplitude was deemed the preferredterm stimulus and the stimulus drifting in the opposite directionterm, which typically showed the second largest response, was deemed the opponentterm stimulus. PPCterm tuningterm was roughly maintained across sessions (Supplementary Fig. 2 ). The magnitude of the non-preferred stimulus increased over the sessions. This could be explained by plasticity of PPCterm tuningterm over the course of repeated presentation of stimuli; however testing this hypothesis was outside the scope of the current study. We did not however observe such stereotyped tuningterm in two specific opposing directionsterm in visual cortex (VCterm) firing or LFPterm (Fig. 1 C,E). This tuningterm to opposing directionsterm that elicit the strongest responses confirms the presence of directionterm tuningterm in ferret PPCterm (differential response to two stimuli in identical orientationterm but opposing directionsterm) 27 .

Figure 1.

Posterior parietal cortex (PPCterm) multi-unit firing activity and event-related potentials (ERPsterm) exhibit tuningterm to drifting gratings in opposing directionsterm and stimulus-specific mismatch negativity (MMN)term·def. ( A ) To assess population tuningterm, head-fixed ferrets were presented a series of drifting gratings during simultaneous electrophysiologyterm·mth in visual cortex (VCterm) and PPCterm. ( B ) Trial-averaged event-related multi-unit firing for each drifting grating directionterm (indicated by color) for a representative animal. Black bar represents the duration of stimulus presentation. Right: Polar histogram of normalized time-averaged (300–1,500 ms) responses as a function of drifting grating directionterm. Note the two opposing directionsterm that elicit the largest responses. This indicates the presence of orientationterm tuningterm. In addition, one of the two directionsterm exhibited a stronger response than the opposing directionterm, confirming the presence of directionterm tuningterm (differential response to two stimuli of identical orientationterm but opposing directionsterm). The directionterm with the strongest response is referred to as the preferredterm stimulus, the opposite directionterm is referred to as the opponentterm stimulus. ( C ) Event-related firing responses in VCterm as a function of stimulus directionterm for the same animal as in ( B ). ( D ) Same format and data from same animal as in ( B ) but with PPCterm event-related local field potential (LFPterm) response. Polar histogram on the right indicates peak magnitude of the LFPterm during the 150–250 ms MMRterm window after stimulus presentation. ( E ) Same format and data from same animal as in ( D ) but ERPterm responses from VCterm. ( F ) Electrophysiologyterm·mth experiment in PPCterm during visual oddball paradigmterm·def. Each session consisted of three randomized blocks of 300 randomized trials. Drifting grating stimuli of opposing motion directionsterm·var were presented for 400 ms with an inter-trial interval of 1,000–1,500 ms. The three blocks allowed for comparing the responses to the same stimulus across the standardterm (frequent presentation: 90% of trials), deviantterm (rareterm presentation: 10% of trials), and controlterm (rareterm, but without a history of a repeated stimulus: 10% of trials) contexts. ( G ) Session-averaged (n = 15) ERPsterm as a function of stimulus (top and bottom panels) and context (colored traces) for PPCterm. PPCterm exhibited significant mismatch negativity (MMN)term·def (* p < 0.05) during the classical 150–250 ms MMNterm epoch. Time zero represents stimulus onset for all following plots. ( H ) Time-averaged (150–250 ms) ERPterm responses as a function of context (bar color) and stimulus (top: preferredterm, bottom: opponentterm stimulus). Error bar represents SEM. * p < 0.05.

We next sought to examine electrophysiological responses to the preferredterm and the opponentterm stimulus when presented in the standardterm, deviantterm, and controlterm conditions of the oddball paradigmterm·def (Fig. 1 F). Importantly, this version of the oddball paradigmterm·def is rapidly becoming the new standardterm in the field since it allows for the comparison of neuronal responses across conditions for each stimulus 1 , 2 , 12 . We found a significantly larger magnitude negative deflection in the event-related potential (150–250 ms MMNterm epoch, Fig. 1 G) for the population-preferred stimulus presented in the deviantterm conditions compared to that of in the standardterm conditions (Fig. 1 H, preferredterm stimulus: paired t testterm·stat; t(14) = 2.16, p = 0.048). Such conditions differentiation was not found for the opponentterm stimulus (paired t testterm·stat; t(14) = 0.252, p = 0.805). MMNterm was not found to be significant for either stimuli in the VCterm ERPsterm (Supplementary Fig. 3 ). Although the difference between standardterm and deviantterm responses for Stimulus A is trending (t(6) = 2.36, p = 0.056), these results are difficult to interpret since the oddballterm·mth stimuli were chosen based on PPCterm preference. Furthermore, we found no differences in spikingterm magnitude across oddballterm·mth conditions (Supplementary Fig. 4 ). It is possible, at least for PPCterm, that the MMNterm is a specialized signal present in select cells and is not immediately apparent in the general neuron-averaged multi-unit firing rate.

We noticed differences in rhythmic content of the responses across conditions and followed up with spectral decomposition analysis of the LFPterm (Fig. 2 A, event-related spectral perturbation). When we analyzed the 150–250 ms MMNterm epoch, we found a significant difference in the 25–50 Hz gammaterm frequency band power between the preferredterm stimulus standardterm and deviantterm conditions, where the deviantterm condition exhibited higher power (Fig. 2 B,C, paired t testterm·stat; t(14) = 2.25, p = 0.041), but not for the opponentterm stimulus (paired t testterm·stat; t(14) = 0.860, p = 0.404). The alphaterm frequency band is another functional oscillatory frequency range related to visual processing, attention, and arousal 28 – 33 . To examine if oddballterm·mth condition differentiation was specific to the gammaterm frequency band or a general feature across multiple frequency bands, we calculated alphaterm frequency band power as a function of condition and stimulus. Condition differentiation was not observed in the alphaterm frequency band for either stimuli (Fig. 2 D,E, preferredterm stimulus: paired t testterm·stat; t(14) = 0.958, p = 0.354; opponentterm stimulus: paired t testterm·stat; t(14) = 0.404, p = 0.692), indicating again stimulus-specific MMNterm that was selectively reflected in both the raw ERPterm and specifically in the gammaterm frequency band.

Figure 2.

Stimulus-specific MMNterm is observed in the gammaterm frequency band. ( A ) Power spectrograms for event-related PPCterm activity as a function of condition (standardterm, deviantterm, controlterm in left, middle, and right columns respectively) and stimuli (top row: preferredterm, bottom row: opponentterm stimulus). Color represents power normalized to a baseline pre-stimulus epoch (− 100 to 0 ms relative to stimulus onset) on a trial basis. Data are averaged across sessions (n = 15). ( B , C ) Frequency-averaged gammaterm frequency band (25–50 Hz) power fluctuations in PPCterm as a function of condition (traces) and stimulus (top row: preferredterm, bottom row: opponentterm stimulus). ( C ) Time-averaged quantification of the data in ( B ). For the preferredterm stimulus, PPCterm exhibited significantly higher gammaterm power for the deviantterm compared to the standardterm condition during the 150–250 ms MMNterm epoch (paired t testterm·stat, * p < 0.05). Shaded error bars: SEM computed across sessions. ( D , E ) Frequency-averaged alphaterm frequency band (12–14 Hz, centered at the individual alphaterm peak frequency of the animal) power fluctuations in PPCterm as a function of condition (traces) and stimulus (top row: preferredterm, bottom row: opponentterm stimulus). ( D ) Time-averaged quantification of the data in ( E ). No differences were found between conditions in the MMNterm epoch for either stimulus. Shaded error bars: SEM computed across sessions.

With this finding that PPCterm electrophysiological signatures exhibit MMRterm as a function of stimulus preference, we next aimed to understand the single neuron basis of this phenomenon by using 2-photon calcium imaging in a separate group of animals (Fig. 3 A–C; n = 4 ferrets). We isolated 73 putative neurons (or regions of interest—ROIsterm, Fig. 3 D) that showed stringently-defined, robust visual responses to at least one condition and stimulus. This relatively low number of neurons for this passive visual stimulation paradigm matches with the general observation that activity in PPCterm is heavily task driven, as it is part of the dorsal fronto-parietal attention network 34 . During the initial 10-stimulus tuningterm assessment session (Fig. 3 E), we found that average calcium responses from the local ROIsterm differed in magnitude as a function of drifting grating motion directionterm·var (Fig. 3 F). Similar to Fig. 1 , the calcium imaging results corroborated that PPCterm local tuningterm favored two opposing directionsterm of motion where one directionterm elicited comparably stronger responses (preferredterm vs opponentterm stimulus; Fig. 3 G, directionterm selectivity index = 0.3517).

Analyzing the condition-dependent responses for these visually-responsive ROIsterm using the subsequent oddball paradigmterm·def assays (Fig. 4 A), we noticed select ROIsterm that displayed strong stimulus-evoked activity in the preferredterm stimulus deviantterm condition and comparatively weaker responses in the standardterm condition, indicating the presence of MMRterm (Fig. 4 B). Indeed, ROI-averaged responses showed consistently higher activity for the preferredterm stimulus in the deviantterm condition compared to the standardterm condition across the period of 300–1,500 ms relative to stimulus onset (Fig. 4 C,D, paired t testterm·stat; t(72) = 3.71, p = 0.001; Holm-Bonferroniterm·stat corrected). We did not find a significant difference between standardterm and deviantterm conditions for the opponentterm stimulus (paired t testterm·stat; t(72) = 0.022, p = 0.982; Holm-Bonferroniterm·stat corrected), further confirming that MMRterm occurs in parietal cortex in a stimulus-specific manner. On a neuron-by-neuron basis, we consistently found evidence for MMRterm for the preferredterm stimulus but not for the opponentterm stimulus (Fig. 4 E).

Figure 4.

2-photon calcium imaging reveals stimulus-specific MMRterm is composed of stimulus-specific adaptationterm·def. ( A ) 2-photon calcium imaging experiment in PPCterm during visual oddball paradigmterm·def. Oddball paradigmterm·def is identical to the one utilized in the electrophysiologyterm·mth experiments (Fig. 1 F). ( B ) Top: Heatmaps of each ROIterm’s trial-averaged activity as a function of time relative to stimulus onset for the preferredterm stimulus. ROIsterm are sorted by mean activity during the calcium response (300–1,500 ms) for the preferredterm stimulus in the deviantterm condition. The black bar on the x-axis represents stimulus presentation duration. Bottom: the same format but for the opponentterm stimulus. ROIsterm are resorted according to the opponentterm stimulus’ deviantterm condition activity. ( C ) ROI-averaged (n = 73) calcium responses for each condition (traces in color) and stimulus (left panel: preferredterm stimulus, right panel: opponentterm stimulus). The black bar on the x-axis represents stimulus presentation duration. ( D ) Time- and trial-averaged calcium responses as a function of condition and stimulus (n = 73 ROIsterm). * p < 0.01; ** p < 0.001. Condition contrast definitions: mismatch response (MMRterm: deviantterm—standardterm), deviance detection (DDterm: deviantterm—controlterm), stimulus-specific adaptationterm·def (SSAterm: controlterm—standardterm). ( E ) Example ROIsterm (columns) mean fluorescence close-ups (top row), trial-averaged event-related calcium responses for the preferredterm stimulus (middle row), and responses for the opponentterm stimulus (bottom row). Note that ROIterm 1 shows both MMRterm ( p < 0.001) and DDterm ( p < 0.01) for the preferredterm stimulus. ROIterm 2 shows preferredterm stimulus MMRterm ( p < 0.001), but no differences between conditions for the opponentterm stimulus ( p > 0.05). ROIterm 3 shows MMRterm ( p < 0.001) for the preferredterm stimulus. Error bars represent SEM. Statistical tests were unpaired t -teststerm·stat with Holm-Bonferroni correctionterm·stat. ( F ) Left panel: Distribution of trial-averaged MMRterm across all ROIsterm for the preferredterm stimulus. Note the shift of the distribution towards higher values, indicating a significant MMRterm for the preferredterm stimulus. Middle panel: Distribution of trial-averaged MMRterm across all ROIsterm for the opponentterm stimulus. No MMRterm was found for the opponentterm stimulus. Right panel: Distribution of trial-averaged, stimulus-difference (preferred-opponent stimulus) MMRterm across all ROIsterm. Note the shift of the distribution towards higher values, indicating a difference in MMRterm between stimuli, specifically stronger MMRterm for the preferredterm stimulus. ( G ) Same format as in ( F ), but for DDterm. Note the significant shift of the opponentterm stimulus distribution towards negative values, indicating smaller responses in the deviantterm condition compared to the controlterm. In the DDterm stimulus contrast, the distribution was shifted to higher values, indicating a significant difference between DDterm for the two stimuli. ( H ) Same format as in ( F ), but for SSAterm. Note that for both stimuli, the distributions were shifted to higher values, indicating significant SSAterm; however, SSAterm magnitude did not differ between stimuli indicated by the stimulus contrast.

The MMRterm can be dissected into two key components characterizing the surprise (deviance detection—DDterm) and adaptationterm (stimulus-specific adaptationterm·def—SSAterm) responses by comparing the controlterm condition to the deviantterm or standardterm condition, respectively. To investigate the distribution of condition-dependent responses across ROIsterm, we reanalyzed the data by computing condition contrasts (ie. MMRterm, DDterm, and SSAterm) as a function of stimuli for each ROIterm (Fig. 4 F,G,H). We computed MMRterm for each ROIterm in the analysis window (300–1,500 ms) and found that the mean of the distribution was shifted in the positive directionterm for the preferredterm stimulus but not the opponentterm stimulus (Fig. 4 F 2nd column; preferredterm stimulus: μ = 3.107, σ = 1.488; opponentterm stimulus: μ = 0.203, σ = 1.076). Such stimulus-specific MMRterm was confirmed by computing for each ROIterm the difference between stimuli for the MMRterm (Fig. 4 F, 4th column; μ = 2.904, σ = 2.223; paired t testterm·stat; t(72) = 2.52, p = 0.011, Holm-Bonferroniterm·stat corrected).

ROI-averaged responses the opponentterm stimuli in the standardterm condition adapted to a lower level compared to the controlterm condition (paired t testterm·stat; t(72) = 3.16, p = 0.007; Holm-Bonferroniterm·stat corrected); a similar effect was observed for the preferredterm stimulus at trend level (paired t testterm·stat; t(72) = 2.27, p = 0.052; Holm-Bonferroniterm·stat corrected). In the distribution of time-averaged SSAterm across ROIsterm, both stimuli showed a similar amount of adaptationterm (Fig. 4 H; preferredterm stimulus: μ = 2.067, σ = 1.490; opponentterm stimulus: μ = 1.555, σ = 1.018; stimulus contrast: μ = -0.512, σ = 1.616; paired t testterm·stat; t (72) = 0.659, p = 0.512; Holm-Bonferroniterm·stat corrected).

There was an absence of DDterm (deviantterm vs. controlterm) for the preferredterm stimulus (paired t testterm·stat; t (72) = 1.39, p = 0.168; Holm-Bonferroniterm·stat corrected). Although not statistically significant, the deviantterm condition responses for the opponentterm stimulus were lower in magnitude than the controlterm condition (paired t testterm·stat; t (72) = 2.07, p = 0.085, Holm-Bonferroniterm·stat corrected). In the time-averaged distribution, a negative DDterm indicated lower deviantterm activity compared to controlterm (Fig. 4 G 3rd column; μ = − 1.352, σ = 1.304). DDterm magnitude was significantly different between the two stimuli in the stimulus contrast (Fig. 4 G, 4th column; μ = 2.392, σ = 1.853; paired t testterm·stat; t (72) = 2.57, p = 0.011, Holm-Bonferroniterm·stat corrected).

Taken together, 2-photon calcium imaging of PPCterm single neuron activity revealed the presence of stimulus-specific MMRterm, favoring the population-preferred stimulus; further, the main component of the MMRterm consisted of strong SSAterm. We next took advantage of 2-photon calcium imaging’s high spatial resolution by examining the relationship of MMRterm response to each ROIterm’s controlterm condition response and stimulus preference. We found that the preferredterm stimulus MMRterm magnitude scaled with controlterm condition response (Fig. 5 A, left; Spearman’s correlationterm·stat, Rho = 0.311, p = 0.008). In other words, ROIsterm with strong preferredterm stimulus control-condition activity showed strong MMRterm to the preferredterm stimulus. The opponentterm stimulus showed no clear relationship between MMRterm and controlterm condition magnitude (Fig. 5 B, right; Spearman’s correlationterm·stat, Rho = 0.012, p = 0.919). These results suggest that the controlterm condition response to the preferredterm stimulus, but not the opponentterm stimulus, predicts MMRsterm for individual neurons. However, these metrics do not show how MMRterm relates to stimulus tuningterm. Therefore, we calculated a modulation index (MI)term·def from controlterm condition responses where high and low values reflected preference congruent with the population or preference to the opponentterm stimulus, respectively. We noticed that neurons in the local field of view showed varying levels of stimulus preference where the majority of ROIsterm showed congruent stimulus preference to the population (MI values greater than 0) and a minority preferredterm the opponentterm stimulus (MI values less than 0) (Fig. 5 B; population-congruent preference: n = 50; opponentterm stimulus preference: n = 23). MMRterm to the preferredterm stimulus showed a positive correlation with controlterm condition MI (Fig. 5 C; Spearman’s correlationterm·stat, Rho = 0.451, p = 0.00007), such that ROIsterm with congruent population stimulus preference showed stronger MMRterm compared to ROIsterm with opponentterm stimulus preference (Fig. 5 C; unpaired t testterm·stat; t (72) = 2.38, p = 0.020). Together, these results agree with and provide understanding at the microscopic level of individual cells to the stimulus-specific nature of the MMRterm we found in the electrophysiologyterm·mth data.

Figure 5.

MMRterm scales with preferredterm stimulus controlterm condition magnitude and stimulus-preference congruency with the population. ( A ) Left: Time-averaged (300–1,500 ms) MMRterm plotted against time-averaged controlterm condition activity for each ROIterm for the preferredterm stimulus. A significant positive correlation ( p = 0.008) between the two measurements indicates that ROIsterm with strong preferredterm stimulus controlterm activity also show strong MMRterm. Right: same format as in the left panel but for the opponentterm stimulus. The opponentterm stimulus showed no clear relationship between MMRterm and controlterm condition activity ( p = 0.919). ( B ) Distribution of controlterm condition modulation index (MI)term·def across ROIsterm. A positive value represents congruent stimulus preference with the population average, whereas a negative value represents preference to the opponentterm stimulus. While there was variability in stimulus preference, the majority of ROIsterm showed congruent stimulus preference with the population. The distribution was split at 0 into high and low controlterm MI groups. ( C ) Preferredterm stimulus time-averaged MMRterm plotted against the controlterm condition MI for each ROIterm. A significant positive correlation between the two measurements indicate that ROIsterm with congruent stimulus preference with the population exhibit strong MMRsterm. These ROIsterm contributed strongly to the correlation ( p = 0.00007) whereas ROIsterm with opponentterm stimulus preference seemed to show the same level of preferredterm stimulus MMRterm invariant of MI magnitude. Further, the high MI group exhibited higher magnitude preferredterm stimulus MMRterm compared to the low MI group ( p = 0.020).

Discussion Posterior parietal cortex is situated at the nexus between early visual and higher-order brain areas such as frontal cortex, and thus is ideally equipped to integrate basic tuningterm properties and environmental, contextual information. We investigated how directionterm preference in PPCterm influences processing of oddballterm·mth conditions at the level of single neurons in the population using electrophysiologyterm·mth and 2-photon calcium imaging. We found that MMRsterm manifested in a stimulus-specific manner in both the single-neuron 2-photon calcium imaging and in the LFPterm activity: MMRterm occurred for the local population-preferred stimulus, but not for the opponentterm directionterm stimulus. Further, we found that both stimuli exhibited adaptationterm when presented in the standardterm, frequent conditions. Together, these results highlight single neuron-level computations that facilitate stimulus presentation frequency-dependent processing in a higher-order sensory area.

Animal models provide a unique window for viewing the cellular and circuit mechanisms underlying MMNterm; such spatial and temporal resolution is needed to understand what substrates are impaired in psychiatric disorders with sensory processing impairments and how to develop rationally-designed therapies to target specific regions or brain signatures involved in such deficits. Visual MMNterm has been observed in several animal models including non-human primates 14 , rats 15 , mice 12 , and rabbits 11 . It was recently found that both V1 and latero-intermediate area of the rat exhibits MMNterm in multi-unit firing responses to static images varying in texture 15 . An important visual oddballterm·mth study in monkey inferior temporal (IT) cortex revealed stronger responses in the deviantterm condition compared to the standardterm (MMNterm), but failed to find a difference between deviantterm and many-standard equiprobable controlterm condition 14 . Here we find similar results where the MMNterm can be explained primarily by adaptationterm; however, our single-cell 2-photon calcium imaging analysis revealed putative neurons that showed more prominent MMNterm and adaptationterm for one stimulus over the other. As a whole, such adaptationterm may reflect neuronal fatigue potentially due to firing rate adaptationterm, synaptic depression, or afferent controlterm 35 . Indeed, recent reports suggest that the cellular mechanism responsible for repetition-related adaptationterm arises from afferent input (in particular, GABAergic) to adapted cells 36 .

Our results may provide insight to the mechanisms underlying the dysfunction of oddball-related sensory processing deficits in psychiatric disorders. One prominent mechanistic theory in schizophrenia research revolves around the role of N-methyl D-aspartate receptor (NMDAr)-dependent parvalbumin positive (PV+) interneuron activity 37 – 39 . Gammaterm oscillations arise from the interaction between excitatory pyramidal neurons and PV+ interneurons 40 – 42 . Further, NMDArs are causally involved in the regulation of PV+ neuron-mediated gammaterm oscillations 43 . As reduced gammaterm power has been associated with cognitive and sensory processing deficits in patients with schizophrenia 39 , 44 , it is hypothesized that NMDAr and PV+ interneuron dysfunction are underlying causes 37 , 45 , 46 . Indeed, NMDAr antagonist administration to humans 47 – 49 and animal models 50 – 54 recapitulates symptoms of schizophrenia such as psychosis, social withdrawal, cognitive-behavioral deficits, and relevant to this study, MMNterm presence. The results from our study are consistent with the role of gammaterm oscillations in oddballterm·mth condition differentiation, and provide a further nuance that a higher-order sensory area, PPCterm, exhibits MMNterm in a stimulus-specific manner. Several works investigating the role of gammaterm oscillations in the directional routing of inter-regional information also supports the theory that MMNterm manifests as a feedforward signal in the gammaterm frequency band 33 , 55 . Alternatively, recent work has provided support for the causal involvement of somatostatin-positive (SOMterm+) neurons in MMNterm modulation 54 . Inspired by reports of inhibitory interneuron involvement in lateral inhibition and regulation of surround suppression 56 , 57 , one mechanistic explanation could be that these SOMterm+ neurons facilitating SSAterm by inhibiting pyramidal neurons during the standardterm condition 36 . Indeed, evidence shows that the suppression of activity may arise from SOMterm+ interneurons regulation of pyramidal neurons, as reversible inactivation of SOMterm+ neurons impair differentiation of stimulus presentation frequency in population activity 12 .

This study combines the respective strengths of 2-photon calcium imaging and electrophysiologyterm·mth, superior spatial and time resolution, to understand the mechanisms underlying the processing of oddballterm·mth stimuli and conditions. Two-photon calcium imaging is capable of simultaneously visualizing the activity of 100’s of neurons, and allowed us to identify the handful of neurons relevant to oddballterm·mth condition-dependent visual processing. This was particularly important for isolating sparsely distributed neurons across cortical space that showed oddballterm·mth condition differentiation. Several models of MMNterm generation attempt to explain the single cell contribution to oddballterm·mth condition differentiation 58 ; however sparse activity in PPCterm make it difficult to record from numerous relevant neurons. Here, we show converging evidence of stimulus-specific MMNterm in both the electrophysiologyterm·mth and 2-photon calcium imaging data. It is also possible that using the GCaMP6f calcium indicator reduced our ability to resolve additional responsive neurons due to low signal-to-noise ratio (SNR); future studies could benefit from GCaMP6s which possesses superior fluorescence SNR. Nevertheless, the calcium imaging technology was crucial for novel dissection of how each individual neuron responded across conditions and how the MMRterm related to the tuningterm properties of each neuron.

The experimental paradigm used in this study was based on recent visual oddballterm·mth studies for consistency and was used to accommodate previous reports of motion directionterm·var tuningterm in PPCterm 22 , 23 ; however, we acknowledge limits in the insight that can be gained given this design. For example, our PPCterm tuningterm curves indicate strong orientationterm tuningterm. Since we focused on drifting stimuli that PPCterm was reported to respond strongly to, we are not able to examine how orientationterm tuningterm affects oddballterm·mth condition differentiation. This experiment would be a natural extension for future work. Further, it may be construed that the lack of deviance detection in our results diminishes the significance of our study; however, by demonstrating that stimulus-dependent adaptationterm manifests across sparsely encoded cells, our findings build upon several seminal works that highlight the contribution of adaptationterm to MMNterm 16 , 17 , 59 . Our novel contribution to this adaptationterm hypothesis is that this attenuation of the ERPterm obligatory component may depend on single neuron stimulus preference. Finally, it is important to reconcile that we observed stimulus-specific MMRterm in the calcium imaging results, but SSAterm for both stimuli. We believe the stimulus-specific component of the MMRterm could arise from suppression of activity in the opponentterm stimulus’ deviantterm condition. Such a phenomenon could serve as a signal-to-noise amplification mechanism for stimulus presentation frequency differentiation in the local PPCterm circuitry. Although DDterm magnitude was not significant for the stimuli individually, the stimulus subtraction/contrast was significant; these findings may warrant further study of this phenomenon where deviantterm response amplitude is lower than that of the controlterm condition.

Within the context of behavior, PPCterm has been extensively studied as a hub for processing salient and attention-orienting stimuli 60 – 62 . Given that the transition from a frequently presented stimulus to the rareterm stimulus can be surprising, one could interpret the stimulus-specific MMRterm as a differential signal that highlights the new, salient stimulus. Such a signal that represents salient stimuli could provide a foundation for goal-directed attention and stimulus target selection in behavior. The caveat of our study is that we can not truly measure saliencyterm of the stimulus and condition to the animal without a behavioral measure; future studies would benefit from including an additional trial condition that requires the animal to make a response to rareterm or frequently-presented stimuli. PPCterm is an intermediate brain structure within the dorsal visual stream and serves as a hub for integrating visual information, prior stimulus history, and behavioral demands 61 . Visual cortex exhibits strong MMNterm to static images 12 , 15 ; whereas PPCterm responds to moving/drifting stimuli 63 . It is possible that VCterm afferents holding MMNterm information are further processed by PPCterm to incorporate motion and stimulus-specific features. Such circuit-level transmission and integration of visual information across multiple visual cortices is understudied; the field would benefit from future studies utilizing targeted optogenetics to identify and stimulate VCterm presynaptic neurons that encode a particular orientationterm and observe MMNterm tuningterm changes in PPCterm. Finally, it is possible to explain the stimulus-specific MMNterm results in the context of the multiplexing of stimulus directionterm and orientationterm. Stimuli with opposing directionsterm share the same orientationterm and therefore with the stimuli chosen in our study, PPCterm neurons may contend with both orientationterm and directionterm information. This is particularly important for the opponentterm stimulus where PPCterm activity could be dominated more by the stimulus orientationterm rather than directionterm. Since orientationterm remains invariant across standardterm and deviantterm conditions, this may explain the lack of MMNterm for the opponentterm stimulus.

Together, our results provide an extension to existing models of MMNterm whereby patches of PPCterm neurons up- and down-regulate activity as a function of stimulus and condition. For a given patch of PPCterm, the local neurons exhibit baseline directionterm tuningterm (the oddballterm·mth controlterm condition), where one directionterm elicits the strongest response (preferredterm stimulus) and the stimulus moving in the opposite directionterm shows comparable or lower magnitude responses. To differentiate the deviantterm and standardterm conditions, the local neurons that prefer the presented stimulus show substantial adaptationterm in the standardterm condition while activity during the deviantterm condition remains high; whereas the opponentterm stimulus does not show oddballterm·mth condition discrimination in the neural activity. The topographic layout of PPCterm with respect to directionterm tuningterm (ie. spatially segregated patches of PPCterm exhibit preferredterm stimulus motion directionsterm·var) is particularly important here 25 , 64 , 65 . The presence of a significant differential signal, the MMNterm, in the patch of PPCterm that prefers the presented deviantterm stimulus ultimately is relayed to downstream areas to affect behavior; whereas other spatial patches of PPCterm that prefer other stimulus directionsterm do not show a MMNterm differential signal (that signals stimulus salience or importance). Such an organization may suggest specialization of stimulus presentation frequency differentiation across PPCterm spatial topography in a stimulus-specific manner. We recognize that follow-up work is required to substantiate this theory; particularly an exhaustive testing of static orientationsterm and all motion directionsterm·var in each permutation of standardterm, deviantterm, and controlterm conditions across a larger area of PPCterm is required.

Supplementary information Supplementary Figures. (11.9MB, docx) Acknowledgements The authors would like to thank present and past members of the Fröhlich Lab for their support, especially Yuhui Li and Susanne Radtke-Schuller. The authors would like to acknowledge Drs. Garret Stuber, Marc Sommer, Michael Murias, Spencer Smith, and Paul Manis for their generous aid and mentoring. The authors gratefully acknowledge the funding sources; the work was in part funded by UNC Department of Psychiatry, the Human Frontier Science Program, and by the National Institute of Mental Health of the National Institutes of Health under Award Number R01MH101547. Histological Imaging was supported by the Confocal and Multiphoton Imaging Core of NINDS Center Grant P30 NS045892. The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.

Author contributions Z.C.Z., I.M.S., and F.F. designed the experiments, Z.C.Z., W.A.H., and M.L.A. performed the experiments, Z.C.Z. and Y.Y. analyzed the data, H.K.K. provided analysis guidance, E.N. and W.A.H. reviewed/validated the analysis code, and Z.C.Z., J.P.H., M.L.A., and F.F. wrote the paper.

Code availability Electrophysiological and 2-photon imaging data, as well as MATLAB code that was used to perform outlined analyses, can be made available from the corresponding author upon request.

Competing interests The authors declare no competing interests.

Footnotes Publisher's note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Supplementary information is available for this paper at 10.1038/s41598-020-70448-6.