Multiple Distinct Timescales of Rapid Sensory Adaption in the Thalamocortical Circuit

Yi Juin Liew; Elaida D. Dimwamwa; Nathaniel C. Wright; Yong Zhang; Garrett B. Stanley

doi:10.1523/JNEUROSCI.1057-24.2025

Multiple Distinct Timescales of Rapid Sensory Adaption in the Thalamocortical Circuit

¹Wallace H Coulter Department of Biomedical Engineering, Georgia Institute of Technology and Emory University, Atlanta, Georgia 30332
²Joint PhD Program in Biomedical Engineering, Georgia Institute of Technology – Emory University – Peking University, Atlanta, Georgia 30332
³Neuroscience Research Institute and Department of Neurobiology, School of Basic Medical Sciences, Peking University, Key Laboratory for Neuroscience, Ministry of Education of China and National Health Commission of the People’s Republic of China, Beijing 100083, China
⁴PKU-IDG/McGovern Institute for Brain Research, Beijing 100871, China

Author contributions: Y.J.L. and G.B.S. designed research; Y.J.L. performed research; Y.J.L. and E.D.D. analyzed data; Y.J.L., E.D.D., N.C.W., Y.Z., and G.B.S. wrote the paper.

Next Section

Abstract

Numerous studies have shown that neuronal representations in sensory pathways are far from static but are instead strongly shaped by the complex properties of the sensory inputs they receive. Adaptation dynamically shapes the neural signaling that underlies our perception of the world yet remains poorly understood. We investigated rapid adaptation across timescales from hundreds of milliseconds to seconds through simultaneous multielectrode recordings from the ventro-posteromedial nucleus of the thalamus (VPm) and layer 4 of the primary somatosensory cortex (S1) in male and female anesthetized mice in response to controlled, persistent whisker stimulation. Observations in VPm and S1 reveal a degree of adaptation that progresses through the pathway. Under these experimental conditions, signatures of two distinct timescales of rapid adaptation in the firing rates of both thalamic and cortical neuronal populations were revealed, also reflected in the synchrony of the thalamic population and in the thalamocortical synaptic efficacy that was measured in putatively monosynaptically connected thalamocortical pairs. Controlled optogenetic activation of VPm further demonstrated that the longer timescale adaptation observed in S1 is likely inherited from slow decreases in thalamic firing rate and synchrony. Despite the degraded sensory responses, adaptation induced by the controlled repetitive stimulation presented here resulted in a shift in coding strategy that favors theoretical discrimination over detection across the observed timescales of adaptation. Overall, although multiple mechanisms contribute to rapid adaptation at distinct timescales, they support a unifying framework on the role of adaptation in sensory processing.

Previous Section Next Section

Significance Statement

Although the perceptual effects of persistent sensory stimulation have been known for centuries, the rapid sensory adaptation of the underlying neural signaling to these persistent inputs are not well understood. Here, we present evidence for two distinct timescales of adaptation over several seconds across the thalamocortical circuit in mice. We identify both the overall level of neural activity and the corresponding population synchrony of the thalamic inputs to primary somatosensory cortex as key role players shaping the cortical adaptation.

Previous Section Next Section

Introduction

The natural sensory environment consists of inputs that span a wide range of spatiotemporal scales. As sensory signals move from the periphery to the cortex, sensory representations are progressively transformed via the characteristics of circuitry at each stage and the nature of the anatomical projections across regions (Maravall et al., 2007, 2013; Adibi and Lampl, 2021). Adaptation has been well documented across all sensory modalities (Whitmire and Stanley, 2016), thought to overcome changes in stimulus statistics via multiple mechanisms that modulate neuronal response properties (Adibi et al., 2013; Latimer et al., 2019). It has long been asserted that the brain's ability to adapt is a fundamental function that is critical for survival, providing a functional advantage by shifting coding strategies in an ethologically advantageous manner (Barlow, 1961). While adaptation has been proposed as a unifying theoretical framework that may integrate numerous disparate observations (Drew and Abbott, 2006), the majority of studies of neuronal adaptation focus on a specific single timescale (Chung et al., 2002; Khatri et al., 2004; Kheradpezhouh et al., 2017). What is not well understood is the evolution of adaptation along different timescales across brain regions in the processing hierarchy and the functional implications.

In the rodent whisker-mediated somatosensory pathway, previous studies have shown that the degree of sensory adaptation increases as neural signals traverse from periphery to the cortex (Ahissar et al., 2000; Khatri et al., 2004; Ganmor et al., 2010; Mohar et al., 2013, 2015). A particularly important stage of the sensory signaling pathway is the thalamocortical circuit, where highly divergent thalamic afferents synapse onto both excitatory and inhibitory neurons in the cortex, initiating thalamic-driven feedforward inhibition that has profound effects on sensory representations (Simons and Carvell, 1989; Gabernet et al., 2005; Cruikshank et al., 2007; Wang et al., 2010a). Thalamocortical synapses exhibit short-term depression, modulating the relative role of the thalamic input (Chung et al., 2002; Castro-Alamancos, 2004a; Katz et al., 2006). However, thalamic neurons themselves also exhibit moderate adaptation, resulting in a withdrawal of the primary drive to the cortex. Recent work also suggests that adaptive changes in thalamic population spiking synchrony play an important role in mediating cortical adaptation through highly timing-sensitive integration in thalamo-recipient cortical neurons (Wang et al., 2010b; Wright et al., 2021). Given the complexity of the thalamocortical circuit, it is likely that multiple mechanisms are engaged in the transformation of sensory representations throughout the pathway. How this myriad of elements collectively evolves across timescales remains unknown.

Here, we investigated in detail the effects of persistent tactile stimuli on neuronal population activity across the thalamocortical circuit and specifically the relative roles of aspects of the thalamic input in shaping the cortical response. We conducted multisite, multielectrode extracellular recordings of spiking activity from the ventro-posteromedial nucleus of the thalamus (VPm) and layer 4 of the whisker response region of primary somatosensory cortex (S1) in anesthetized mice. We presented controlled, repetitive, punctate whisker deflection trains (300°/s sawtooth deflections at 10 Hz; see Materials and Methods) to individual whiskers and quantified the adaptive effects on single-unit spiking activity in VPm and putative excitatory and inhibitory S1 neurons, as well as in a subset of putative synaptically connected thalamocortical pairs. Under these experimental conditions, in both the cortex and the thalamus, we uncovered evidence of two distinct timescales in the rapid adaptation dynamics over a timecourse of seconds that exhibited differential functional properties. We observed an evolution in the relative roles of individual mechanisms across these timescales, with cortical adaptation at the longer timescale appearing to primarily reflect the adaptation of thalamic drive and synchrony. As a result, from the perspective of an ideal observer of spiking activity, adaptation in both of the distinct timescales also continually shifts toward better discrimination of whisker deflection velocity at the expense of detection. These results collectively suggest a complex interplay of adaptive elements that evolve on distinct timescales and position thalamic population activity as a key role player in shaping cortical representations.

Previous Section Next Section

Materials and Methods

Animals

Experiments were conducted using 10 adult C57BL/6J and 14 adult NR133 × Ai32 (NR133 Cre-recombinase driver line crossed with Ai32 mice for expression of channelrhodopsin in the VPm; Gerfen et al., 2013; Bolus et al., 2021) mice of both sexes (14 male, 10 female) aged between 11 and 44 weeks (mean: 20 weeks). All procedures were approved by the Georgia Institute of Technology Institutional Animal Care and Use Committee and followed guidelines established by the National Institutes of Health.

Head-plate implantation

Head-plate implantation was performed 3–7 d before electrophysiological recordings were conducted. A lightweight custom metal (titanium or stainless steel) head-plate was implanted on the skull of the mouse for head fixation and improved stability during recording, in accordance with a previously described protocol (Borden et al., 2017; Liew et al., 2021; Wright et al., 2021; Dimwamwa et al., 2023). During this survival surgical procedure, the animal was sedated with 5% vaporized isoflurane, and anesthesia was maintained with 2–3% isoflurane for the remainder of the procedure. An opioid and nonsteroidal anti-inflammatory analgesic were administered postoperatively (SR-buprenorphine 0.8–1 mg/kg, s.c., and ketoprofen 5–10 mg/kg, i.p.). The animal's eyes were covered with ophthalmic eye ointment (Puralube Vet Ointment) to prevent dehydration. Body temperature was monitored and maintained at 37°C. After sterilizing the skin above the skull and applying topical anesthetic (lidocaine 2%, 0.05 ml, max 7 mg/kg, s.c.), we removed the skin and scraped off periosteum and any conjunctive tissue from the skull. We gently separated tissue and muscles close to the lateral edge of the skull using a scalpel blade (Glass Van & Technocut, size no. 15), leaving sufficient room for head-plate attachment, away from the targeted recording areas. Especially critical for targeting of the ventro-posteromedial nucleus of the thalamus (VPm), the skull surface was leveled during head fixation by adjusting the animal's head position to minimize the relative height difference (<100 µm) between bregma and lambda (Franklin and Paxinos, 2008). The metal head-plate was secured over the intact skull with C&B Metabond (Parkell Prod), and the rest of the attachment site and edges of the skin incision were secured with skin adhesive (Vetbond, 3 M). The final head-plate and dental acrylic structure created a well for the containment of Ringer's solution (in mM: 135 NaCl, 5 KCl, 5 HEPES, 1 MgCl₂-6H₂O, 1.8 CaCl₂-2H₂Ol, pH 7.3) for future imaging and electrophysiological recording sessions. We applied a thin layer of transparent Metabond dental acrylic over both hemispheres to prevent inflammation and covered the exposed skull with a silicone elastomer (Kwik-Cast, World Precision Instruments).

Intrinsic signal optical imaging of the mouse S1

Intrinsic signal optical imaging of whisker evoked responses in mouse primary somatosensory cortex was performed through the skull under 1–1.2% isoflurane anesthesia to functionally generate a map of the precise location of S1 barrel columns (Lefort et al., 2009; Pala and Petersen, 2015). The imaging was conducted in a separate session to the anesthetized electrophysiology, either within the same session as head-plate implantation or in a separate session. The skull was thinned using a dental drill (0.1-mm-diameter bit, Komet) until blood vessels became visible under warm Ringer's solution. The Ringer's was then covered with a glass coverslip (thickness 0.13–0.17 mm, Fisherbrand). A reference image of the blood vasculature was captured under green illumination (530 nm; M530F1 LED, Thorlabs). Repetitive whisker stimuli (4 s of stimuli at 10 Hz; sawtooth waveform; 1,000°/s) were delivered to individual whiskers (see below, Sensory stimulus) while imaging S1 under red illumination (625 nm; M625F1 LED, Thorlabs). Images were acquired using a CCD Camera (MiCam02HR, SciMedia) at 100 Hz. Using previously published methods, we recorded intrinsic signals from at least three barrel columns and estimated the location of the unmapped columns by overlaying a template S1 barrel map reconstructed from histology (Liew et al., 2021). At the end of the imaging session, we applied transparent Metabond dental acrylic directly on the S1 site to protect the thinned skull from contamination and infection and sealed the exposed skull with a silicone elastomer (Kwik-Cast, World Precision Instruments).

Anesthetized electrophysiology

Animals were sedated with 5% vaporized isoflurane and maintained with 2–3% isoflurane. Their eyes were covered ophthalmic eye ointment and the silicone elastomer removed from the skull surface. The skull above the VPm and S1 was carefully thinned with a dental drill (0.1-mm-diameter bit, Komet), while frequently irrigating with Ringer's solution frequently to remove debris and prevent overheating. When the skull was thin enough to easily puncture, we made small craniotomies above VPm and S1. The craniotomies were constantly irrigated with Ringer's solution to maintain the integrity of the brain tissue.

The animals were then maintained with 1–2% isoflurane for the electrophysiological recordings. Either in separate sessions or in the same session for paired recordings, 32-channel silicon probes with the “Poly-3” configuration (A1×32-Poly3-5mm-25s-177, A1×32-Poly3-6mm-50s-177, or A1×32-Poly3-10mm-50s-177, NeuroNexus) were independently lowered into the VPm and S1. All signals were acquired using the Tucker-Davis Technologies RZ2 Bioprocessor. Neuronal signals were amplified, bandpass filtered (500 Hz–5 kHz), and digitized at 24,414.0625 Hz per channel. Stimulus waveforms and other continuous data were digitized at 10 kHz per channel.

VPm recordings

For VPm recordings, the electrode was inserted perpendicular to the cortical surface and typically positioned 1.8 mm caudal and 1.8 mm lateral to bregma and lowered to a depth of 3–3.5 mm. If VPm could not be functionally located using this method, we also targeted VPm using coordinates 1.0–1.2 mm from the B1 or Beta columns of S1 as identified via the intrinsic signal optical imaging map.

S1 recordings

For S1 recordings, a small craniotomy was created at the desired location based on the intrinsic signal optical imaging map, being careful to remain distant from blood vessels. The electrode was positioned at a 35° angle from the vertical axis (parallel to S1) and cortical neurons were recorded at a stereotaxic depth of 300–600 µm from the cortical surface, corresponding to layer 4 (L4).

Barreloid/barrel targeting

Any barreloid and/or barrel corresponding to whiskers in the first 4 columns (1–4) and any row (A–E) were considered for recordings, and multiple recordings were conducted per mouse to target multiple barreloids/barrels. First, candidate principal whiskers (PWs), or the whisker that elicits the earliest and largest response for a given barreloid/barrel were identified through manual stimulation. Then, the responses to various sensory stimuli (see below, Sensory stimulus) were methodically recorded during computer-controlled galvo-motor stimulation for 3–4 separate whiskers and the principal whisker for the recorded neurons was quantitatively determined in postprocessing (see below, Principal whisker (PW) identification).

After each recording session, the probes were removed, exposed tissue was covered with agarose, and the well was sealed with Kwik-Cast. Subsequent recording sessions were done every other day. One recording session was conducted per day and two to three recording sessions were conducted for each mouse using either the original craniotomy or additional craniotomies targeting a different cortical column and barreloid.

Sensory stimulus

Controlled, single whisker stimulation was delivered to the identified whiskers of interest in the rostral-caudal plane using a computer-controlled galvo-motor (galvanometer optical scanner model 6210H, Cambridge Technology). Whiskers were trimmed ∼12 mm from the face. The whisker was fed into the insertion hole at the end of an extension tube (inner diameter: ∼200–300 µm; length, 15 mm) that was connected to the rotor of the galvanometer stimulator. The stimulus probe was positioned 10 mm from the vibrissal pad. The range of motion of the galvanometer was ±20°, with a bandwidth of 200 Hz. The galvanometer system was controlled using a custom developed hardware/software system (MATLAB Real-Time Simulink System, MathWorks or the Real- time experiment Interface application; http://rtxi.org/), with a 1 kHz sampling rate. The intertrial interval (time between end of one trial and beginning of the next) for all stimuli was at least 15 s, to minimize any residual effects from the previous trial.

Adapting stimulus

To induce rapid sensory adaptation, a train of punctate whisker deflections was presented at 10 Hz (i.e., 100 ms interdeflection interval) for 10 s in the rostral-caudal plane using the galvanometer as described above. Each punctate deflection within the adapting train consisted of a 300°/s sawtooth waveform with an exponential rise and decay lasting 17 ms. The reported velocity is defined by the average velocity over the 8.5 ms rising phase (Wang et al., 2010b). The neuronal sensory response for control, 1.5 s adapted, or 10 s adapted was then quantified for the 1st, 15th, or 100th stimulus in the train.

For the assessment of velocity sensitivity as well as the detectability and discriminability from the perspective of an ideal observer, the response to probe stimuli is what is analyzed. Probe stimuli consist of single sawtooth deflections delivered at one of five velocities: 50, 300, 450, 600, and 1,200°/s. The probe deflections were presented with either no preceding adapting stimulus (Control) or with a preceding 1.5 s adapting stimulus (15 deflections at 10 Hz) or 10 s adapting stimulus (100 deflections at 10 Hz). The probe was delivered 100 ms following the end of the adapting trains. All stimulus conditions were randomized across trials, and at least 20 trials per stimulus condition were typically obtained for different velocities.

Weak sinusoidal stimulus

To assess monosynaptic connectivity between VPm and S1 neurons, the thalamocortical circuit was probed using a weak, desynchronizing sinusoidal whisker stimulus to elevate baseline firing rates in recorded neurons (Liew et al., 2021). Specifically, 2–4 Hz weak sinusoidal deflections of 4° amplitude and 4 s duration (Bruno and Simons, 2002; Bruno and Sakmann, 2006; Wang et al., 2010b) were delivered for ∼200–500 trials to obtain at least 2,000 spikes for the cross-correlation analysis. Occasionally, the first 0.5–1 s of the trials were eliminated due to high firing rates at the onset of stimulus presentation.

Optogenetic stimulation of neural activity

A subset of recordings was performed on NR133 × Ai32 transgenic mice to directly activate VPm neurons, bypassing sensory activation from the periphery. In these experiments, a 32-channel silicon opto-electrode with an attached 105 µm fiber (A1×32-Poly3-5mm-25s-177-OA32LP NeuroNexus) coupled to a 200 µm optic fiber (M95L01, Thorlabs) and a 470 nm LED (M470F3, Thorlabs) was lowered into VPm. The optogenetic stimulation was designed to emulate the adaptation induced by the sensory stimulation described above. The optogenetic adapting stimulus consisted of 470 nm LED pulses that were 17 ms in duration and were delivered to VPm cell bodies for 10 s at 10 Hz.

Postmortem histology

To verify the recording sites in VPm and S1, a subset of animals was perfused after the paired recording experiments. To label each probe recording track, the probe was slowly retracted along its axis of entry at the conclusion of recording and a few drops of DiI (2.5 mg/ml, in ethanol) were applied on the probe. The electrode was then inserted into the same penetration site along the same axis, and the probe was left in the brain for at least 15 min. The animal was then killed with an overdose of sodium pentobarbital (euthasol, 0.1 ml at 390 mg/ml), a transcardial perfusion was performed, the brain was extracted, and the brain was fixed in 4% paraformaldehyde (PFA, Electron Microscopy Sciences) overnight. The brain was sliced in 100 µm coronal sections. The slices were incubated with DAPI (2 mM in PBS, AppliChem) for 15 min before being mounted on slides with a DABCO solution (1,4-diazabicyclo[2.2.2]octane, Sigma). The slices were imaged with a confocal microscope (Laser Scanning Microscope 900, Zeiss) to verify the overlap between the DiI stained recording track and the CO-stain.

Analytical methods

Spike sorting

Spike sorting was performed offline using KiloSort2.5 (Pachitariu et al., 2024) followed by manual curation of neuronal clusters in Phy2 (Rossant and Harris, 2013). Well-isolated single units were classified using the following criteria: mean waveform signal-to-noise ratio >3 and <1 and 3% of all spikes violating the 2 ms refractory period for cortical and VPm neurons, respectively (Liew et al., 2021; Dimwamwa et al., 2023).

Analysis of sensory response

For each neuron, only the whisker response to the identified PW was analyzed (see below, Principal whisker (PW) identification). Whisker evoked activity was quantified within the neural response window for each neuron, defined as a 30 ms window following stimulus onset. Evoked response amplitudes were characterized from the trial-to-trial spike count in the neural response window following each deflection. When applicable, evoked responses were normalized to the maximum response to the first three stimuli. The normalized evoked response amplitude was defined as the spike count in the response window divided by the maximum response amplitude evoked by the first three deflections in the adapting trains. The adaptation ratio was computed for each neuron as the average spike count elicited by the last three deflections, normalized to the averaged response to the first three deflections. First-spike latencies were defined as the average time of the first spike within the response window across trials in which there was a spike within the 30 ms response widow (Storchi et al., 2012).

VPm neuron validation

It is important to take great care to distinguish different thalamic nuclei from one another, particularly in the mouse where the regions are very small and close together. Thus, all analyses only considered neurons that were responsive to the deflection of at least one whisker. Whisker responsiveness was determined by performing the Wilcoxon signed-rank test on the poststimulus spiking in the 30 ms neuronal response window compared with an equivalent duration prestimulus window. Then, a combination of measures was used to classify whisker responsive neurons as originating from the VPm. First, for isolated (nonadapting) stimuli, we quantified the first-spike latency. Second, we quantified the shift in first-spike latency response to adapting stimuli, defined as the time difference between the first-spike latency to the first and last adapting stimuli (Masri et al., 2008). Third, we calculated the response reliability, defined as the percentage of trials where a response was detected within 20 ms in response to each deflection of the repetitive stimulation (8–10 Hz; Mainen and Sejnowski, 1995). Only neurons with an average first-spike latency of <12 ms, latency shift of <20 ms, and response reliability of >20% were included (Liew et al., 2021; Wright et al., 2021; Borden et al., 2022; Dimwamwa et al., 2023).

Cortical neuron classification

S1 layer 4 is the major thalamocortical recipient of VPm projections, where a diversity of cell types receive the feedforward synaptic inputs (Bruno and Simons, 2002). Here, cortical neurons were classified as regular-spiking (RS) or fast-spiking (FS) based on the time interval between the trough and peak of the mean spike waveform, with a cutoff of 0.4 ms separating the broad RS neurons from the narrow FS neurons (Barthó et al., 2004).

Principal whisker (PW) identification

Of the 3–4 whiskers chosen for computer-controlled galvo-motor stimulation, the PW was assigned based on the whisker that elicited the largest peak and smallest latency sensory response in the 30 ms neuronal response window. Each VPm neuron was individually assigned a PW using the LFP response for the channel to which the neuron belonged. All simultaneously recorded S1 neurons were collectively assigned a PW based on the MU response across all channels. The sensory responses presented throughout the manuscript include only trials in which each neuron's primary whisker was being deflected.

Determining topographic alignment

A VPm→S1 neuron pair was verified to be topographically aligned when they shared the whisker that elicits the maximum response to isolated stimuli (both spike evoked response and LFP amplitude) and the mean spike latency difference in the sensory response between VPm and S1 neuron pair was between 1 and 5 ms.

Characterization of the adaptation dynamics with exponential models

The adaptive reduction in response amplitudes was characterized by either single or bi-exponential models applied to the normalized evoked response or synchrony over the 10 s adapting stimulus period for the response of either the mean population or individual neurons. Normalization was to the maximum response/synchrony for the first three stimuli.

For the single exponential fit, the following exponential function was used:<mml:math display="block"><mml:mi>y</mml:mi><mml:mo stretchy="false">(</mml:mo><mml:mi>t</mml:mi><mml:mo stretchy="false">)</mml:mo><mml:mo>=</mml:mo><mml:mi>a</mml:mi><mml:msup><mml:mi>e</mml:mi><mml:mrow><mml:mo>−</mml:mo><mml:mo stretchy="false">(</mml:mo><mml:mi>t</mml:mi><mml:mo>/</mml:mo><mml:mi>τ</mml:mi><mml:mo stretchy="false">)</mml:mo></mml:mrow></mml:msup><mml:mo>+</mml:mo><mml:mi>b</mml:mi><mml:mo>,</mml:mo></mml:math>y(t)=ae−(t/τ)+b, where <mml:math display="inline"><mml:mi>y</mml:mi><mml:mo stretchy="false">(</mml:mo><mml:mi>t</mml:mi><mml:mo stretchy="false">)</mml:mo></mml:math>y(t) is the normalized evoked response at time t, τ is the time constant of adaptation, a is a scaling coefficient, and b is a constant representing an offset. For the bi-exponential fit, the following function was used:<mml:math display="block"><mml:mi>y</mml:mi><mml:mo stretchy="false">(</mml:mo><mml:mi>t</mml:mi><mml:mo stretchy="false">)</mml:mo><mml:mo>=</mml:mo><mml:msub><mml:mi>a</mml:mi><mml:mn>1</mml:mn></mml:msub><mml:msup><mml:mi>e</mml:mi><mml:mrow><mml:mo>−</mml:mo><mml:mo stretchy="false">(</mml:mo><mml:mi>t</mml:mi><mml:mo>/</mml:mo><mml:msub><mml:mi>τ</mml:mi><mml:mn>1</mml:mn></mml:msub><mml:mo stretchy="false">)</mml:mo></mml:mrow></mml:msup><mml:mspace width="0.25em"></mml:mspace><mml:mo>+</mml:mo><mml:mspace width="0.25em"></mml:mspace><mml:msub><mml:mi>a</mml:mi><mml:mn>2</mml:mn></mml:msub><mml:msup><mml:mi>e</mml:mi><mml:mrow><mml:mo>−</mml:mo><mml:mo stretchy="false">(</mml:mo><mml:mi>t</mml:mi><mml:mo>/</mml:mo><mml:msub><mml:mi>τ</mml:mi><mml:mn>2</mml:mn></mml:msub><mml:mo stretchy="false">)</mml:mo></mml:mrow></mml:msup><mml:mspace width="0.25em"></mml:mspace><mml:mo>+</mml:mo><mml:mi>b</mml:mi><mml:mo>,</mml:mo></mml:math>y(t)=a1e−(t/τ1)+a2e−(t/τ2)+b, where <mml:math display="inline"><mml:mi>y</mml:mi><mml:mo stretchy="false">(</mml:mo><mml:mi>t</mml:mi><mml:mo stretchy="false">)</mml:mo></mml:math>y(t) is the normalized evoked response at time t, <mml:math display="inline"><mml:msub><mml:mi>τ</mml:mi><mml:mrow><mml:mn>1</mml:mn><mml:mspace width="0.25em"></mml:mspace></mml:mrow></mml:msub></mml:math>τ1 and <mml:math display="inline"><mml:msub><mml:mi>τ</mml:mi><mml:mn>2</mml:mn></mml:msub></mml:math>τ2 are the time constants of two timescales of adaptation, <mml:math display="inline"><mml:msub><mml:mi>a</mml:mi><mml:mrow><mml:mn>1</mml:mn><mml:mspace width="0.25em"></mml:mspace></mml:mrow></mml:msub></mml:math>a1 and <mml:math display="inline"><mml:msub><mml:mi>a</mml:mi><mml:mn>2</mml:mn></mml:msub></mml:math>a2 are the scaling coefficients, and b is a constant representing an offset. For the tri-exponential fit, the following function was used:<mml:math display="block"><mml:mi>y</mml:mi><mml:mo stretchy="false">(</mml:mo><mml:mi>t</mml:mi><mml:mo stretchy="false">)</mml:mo><mml:mo>=</mml:mo><mml:msub><mml:mi>a</mml:mi><mml:mn>1</mml:mn></mml:msub><mml:msup><mml:mi>e</mml:mi><mml:mrow><mml:mo>−</mml:mo><mml:mo stretchy="false">(</mml:mo><mml:mi>t</mml:mi><mml:mo>/</mml:mo><mml:msub><mml:mi>τ</mml:mi><mml:mn>1</mml:mn></mml:msub><mml:mo stretchy="false">)</mml:mo></mml:mrow></mml:msup><mml:mspace width="0.25em"></mml:mspace><mml:mo>+</mml:mo><mml:mspace width="0.25em"></mml:mspace><mml:msub><mml:mi>a</mml:mi><mml:mn>2</mml:mn></mml:msub><mml:msup><mml:mi>e</mml:mi><mml:mrow><mml:mo>−</mml:mo><mml:mo stretchy="false">(</mml:mo><mml:mi>t</mml:mi><mml:mo>/</mml:mo><mml:msub><mml:mi>τ</mml:mi><mml:mn>2</mml:mn></mml:msub><mml:mo stretchy="false">)</mml:mo></mml:mrow></mml:msup><mml:mspace width="0.25em"></mml:mspace><mml:mo>+</mml:mo><mml:msub><mml:mi>a</mml:mi><mml:mn>3</mml:mn></mml:msub><mml:msup><mml:mi>e</mml:mi><mml:mrow><mml:mo>−</mml:mo><mml:mo stretchy="false">(</mml:mo><mml:mi>t</mml:mi><mml:mo>/</mml:mo><mml:msub><mml:mi>τ</mml:mi><mml:mn>3</mml:mn></mml:msub><mml:mo stretchy="false">)</mml:mo></mml:mrow></mml:msup><mml:mspace width="0.25em"></mml:mspace><mml:mo>+</mml:mo><mml:mspace width="0.25em"></mml:mspace><mml:mi>b</mml:mi><mml:mo>,</mml:mo></mml:math>y(t)=a1e−(t/τ1)+a2e−(t/τ2)+a3e−(t/τ3)+b, where <mml:math display="inline"><mml:mi>y</mml:mi><mml:mo stretchy="false">(</mml:mo><mml:mi>t</mml:mi><mml:mo stretchy="false">)</mml:mo></mml:math>y(t) is the normalized evoked response at time t; <mml:math display="inline"><mml:msub><mml:mi>τ</mml:mi><mml:mrow><mml:mn>1</mml:mn><mml:mspace width="0.25em"></mml:mspace><mml:mo>,</mml:mo></mml:mrow></mml:msub><mml:msub><mml:mi>τ</mml:mi><mml:mrow><mml:mn>2</mml:mn><mml:mo>,</mml:mo></mml:mrow></mml:msub></mml:math>τ1,τ2, and <mml:math display="inline"><mml:msub><mml:mi>τ</mml:mi><mml:mn>3</mml:mn></mml:msub></mml:math>τ3 are the time constants of three timescales of adaptation; <mml:math display="inline"><mml:msub><mml:mi>a</mml:mi><mml:mn>1</mml:mn></mml:msub></mml:math>a1 , <mml:math display="inline"><mml:msub><mml:mi>a</mml:mi><mml:mn>2</mml:mn></mml:msub></mml:math>a2 , and <mml:math display="inline"><mml:msub><mml:mi>a</mml:mi><mml:mn>3</mml:mn></mml:msub></mml:math>a3 are scaling coefficients; and b is a constant representing an offset (Patterson et al., 2013). All parameters were optimized using the “fmincon” function in MATLAB. The optimization aimed to minimize the sum of squared differences between the observed data points and the model predictions. The time constants, the offset, and scaling coefficients were constrained to be non-negative. We additionally imposed a constraint requiring that subsequent time constants be at least <mml:math display="inline"><mml:msqrt><mml:mn>10</mml:mn></mml:msqrt></mml:math>10 times larger than the previous time constant, empirically determined to separate timescales. Omission of this constraint does not change the fundamental finding of best fits with bi-exponential models. Note that for the multi-exponentials, the exponents were sorted in order from fastest to slowest time constant.

Importantly, the extra degree of freedom in bi- and tri-exponential functions could result in an improvement in model fit regardless of the underlying structure in the data. Thus, the goodness of fit of each model was quantified using the Akaike information criterion (AIC) which estimates the prediction error while accounting for increases in model complexity. The AIC penalizes extra model parameters and thus serves to account for overfitting. The different model complexities were assessed by comparing the AIC for each model across the dataset, with the smallest AIC indicating the best fit. The AIC was computed for each model according to the following equation:<mml:math display="block"><mml:mrow><mml:mi mathvariant="normal">AIC</mml:mi></mml:mrow><mml:mo>=</mml:mo><mml:mn>2</mml:mn><mml:mi>k</mml:mi><mml:mo>−</mml:mo><mml:mn>2</mml:mn><mml:mrow><mml:mi mathvariant="normal">ln</mml:mi></mml:mrow><mml:mo stretchy="false">(</mml:mo><mml:mrow><mml:mrow><mml:mover><mml:mi>L</mml:mi><mml:mo stretchy="false">^</mml:mo></mml:mover></mml:mrow></mml:mrow><mml:mo stretchy="false">)</mml:mo><mml:mo>,</mml:mo></mml:math>AIC=2k−2ln(L^), where k is the number of model parameters to fit and <mml:math display="inline"><mml:mrow><mml:mover><mml:mi>L</mml:mi><mml:mo stretchy="false">^</mml:mo></mml:mover></mml:mrow></mml:math>L^ is the log likelihood function for the model (Ljung, 1987). The log likelihood function is assessed by assuming normally distributed errors:<mml:math display="block"><mml:mrow><mml:mover><mml:mi>L</mml:mi><mml:mo stretchy="false">^</mml:mo></mml:mover></mml:mrow><mml:mo>=</mml:mo><mml:mo>−</mml:mo><mml:mstyle displaystyle="true" scriptlevel="0"><mml:mrow><mml:mfrac><mml:mi>n</mml:mi><mml:mn>2</mml:mn></mml:mfrac></mml:mrow><mml:mi>log</mml:mi><mml:mspace width="0.2em"></mml:mspace><mml:mo stretchy="false">(</mml:mo><mml:mrow><mml:mn>2</mml:mn><mml:mi>π</mml:mi></mml:mrow><mml:mo stretchy="false">)</mml:mo><mml:mo>−</mml:mo><mml:mstyle displaystyle="true" scriptlevel="0"><mml:mrow><mml:mfrac><mml:mi>n</mml:mi><mml:mn>2</mml:mn></mml:mfrac></mml:mrow><mml:mi>log</mml:mi><mml:mspace width="0.2em"></mml:mspace><mml:mo stretchy="false">(</mml:mo><mml:mrow><mml:msup><mml:mi>σ</mml:mi><mml:mn>2</mml:mn></mml:msup></mml:mrow><mml:mo stretchy="false">)</mml:mo><mml:mo>−</mml:mo><mml:mspace width="0.25em"></mml:mspace><mml:mstyle displaystyle="true" scriptlevel="0"><mml:mrow><mml:mfrac><mml:mn>1</mml:mn><mml:mrow><mml:mn>2</mml:mn><mml:msup><mml:mi>σ</mml:mi><mml:mn>2</mml:mn></mml:msup></mml:mrow></mml:mfrac></mml:mrow><mml:munderover><mml:mrow><mml:mo movablelimits="false">∑</mml:mo></mml:mrow><mml:mrow><mml:mi>i</mml:mi><mml:mo>=</mml:mo><mml:mn>1</mml:mn></mml:mrow><mml:mi>n</mml:mi></mml:munderover><mml:mspace width="0.2em"></mml:mspace><mml:mo stretchy="false">(</mml:mo><mml:msub><mml:mi>y</mml:mi><mml:mi>i</mml:mi></mml:msub><mml:mo>−</mml:mo><mml:msub><mml:mrow><mml:mover><mml:mi>y</mml:mi><mml:mo stretchy="false">^</mml:mo></mml:mover></mml:mrow><mml:mi>i</mml:mi></mml:msub><mml:msup><mml:mo stretchy="false">)</mml:mo><mml:mn>2</mml:mn></mml:msup><mml:mo>,</mml:mo></mml:mstyle></mml:mstyle></mml:mstyle></mml:math>L^=−n2log(2π)−n2log(σ2)−12σ2∑i=1n(yi−y^i)2, where n is the number of observations (100), <mml:math display="inline"><mml:msub><mml:mi>y</mml:mi><mml:mi>i</mml:mi></mml:msub></mml:math>yi is the observed response for the <mml:math display="inline"><mml:msup><mml:mi>i</mml:mi><mml:mrow><mml:mi>t</mml:mi><mml:mi>h</mml:mi></mml:mrow></mml:msup></mml:math>ith stimulus in the train, and <mml:math display="inline"><mml:msub><mml:mrow><mml:mover><mml:mi>y</mml:mi><mml:mo stretchy="false">^</mml:mo></mml:mover></mml:mrow><mml:mi>i</mml:mi></mml:msub></mml:math>y^i is the model estimate for the response to the <mml:math display="inline"><mml:msup><mml:mi>i</mml:mi><mml:mrow><mml:mi>t</mml:mi><mml:mi>h</mml:mi></mml:mrow></mml:msup></mml:math>ith stimulus in the train and <mml:math display="inline"><mml:msup><mml:mi>σ</mml:mi><mml:mn>2</mml:mn></mml:msup></mml:math>σ2 is the variance of the error.

To better visualize the exponential decay of the adaptation response, we examined the data within a semi-log plot. We overlaid linear regression fits on each segment of the data to highlight the two clear regimes of decay, further highlighting the bi-exponential nature of the sensory timecourses and the single exponential nature of the optogenetic timecourses.

Analysis of velocity sensitivity

To measure the sensitivity of thalamocortical neurons to sensory stimulus velocity, the whisker response was quantified as a function of increasing deflection velocities: 50, 125, 300, 600, 900, 1,200°/s. These stimuli were presented either in isolation (Control) or immediately following adapting stimuli (1.5 s or 10 s), all randomly interleaved.

Spike cross-correlation analysis

Spike cross-correlation analysis was performed to infer monosynaptic connectivity and compute synaptic efficacy for paired recordings in topographically aligned thalamocortical regions, as well as to compute VPm population synchrony. For each spike in a “reference” neuron, a cross-correlogram (CCG) histogram of the relative times of the spikes from a “target” neuron was generated within a 25 ms window before and after each reference spike. All CCGs were constructed using 0.5 ms bins.

Inferring monosynaptic connectivity

Using a previously developed method for monosynaptic connectivity inference, in paired and topographically aligned recordings, the thalamocortical circuit was externally driven with a weak whisker stimulus (see above, Weak sinusoidal stimulus; Liew et al., 2021). For each topographically aligned VPm→S1 pair, a CCG of the S1 neuron's spiking relative to the VPm neurons spiking during the weak whisker stimulus was constructed, referred to as the raw CCG. To correct for correlated, stimulus-locked spiking, a shuffle-corrected (SC) spike cross-correlogram was generated. Specifically, a shuffled CCG was generated by randomly shuffling the trials of the VPm and S1 neuron spiking activity relative to each other. The SC CCG was then generated by subtracting the shuffled CCG from the raw CCG. This effectively corrects for any elements of the CCG that are due to the stimulus. To conclude that a neuronal pair was putatively “connected,” two criteria were adopted that were expanded from previous studies (Reid and Alonso, 1995; Swadlow and Gusev, 2001; Bruno and Sakmann, 2006) based on both the raw and shuffle-corrected cross-correlogram: (Criterion 1) a notable sharp, millisecond-fine peak observed within a narrow lag of 1–4 ms after a VPm spike that is the maximum in the raw CCG; and (Criterion 2) this fast “monosynaptic peak” is significant or still present after accounting for (subtracting) stimulus-induced correlations in the SC CCG (Liew et al., 2021). Because this approach is statistical in nature, and not definitive, we refer to such pairs as “putative” monosynaptically connected pairs.

Measuring synaptic efficacy

To quantify the effects of adaptation on synaptic transmission across the thalamocortical junction, we measured synaptic efficacy, quantified as the number of spikes within a 1–3 ms window of the SC CCG divided by the total number of VPm spikes used in the analysis (Swadlow and Gusev, 2001). Synaptic efficacy was computed per stimulus in the adapting train using an 80 ms poststimulus response window.

Measuring VPm synchrony

To measure the level of synchronization in the VPm, synchrony was computed for all pairs of simultaneously recorded VPm neurons that responded to the same principal whisker. Synchrony was quantified using the total number of spikes within the central ±7.5 ms window of the CCG generated from stimulus-evoked spiking in the 30 ms response window divided by the total number of spikes in the raw CCG (Borden et al., 2017; Wright et al., 2021). When applicable, normalized synchrony was normalized to the maximum response to the first three stimuli.

Receiver operator characteristic (ROC) analysis

To evaluate the functional consequences of the effects of adaptation over timescales, the theoretical detectability of sensory features was assessed for significantly responsive RS and FS neurons by applying ideal observer analysis (Wang et al., 2010b; Ollerenshaw et al., 2014; Whitmire and Stanley, 2016). The noise distribution was generated from baseline activity, or the spike count 30 ms before the stimulus onset in each trial. For each cortical neuron response to a stimulus at a given velocity, the signal came from a probability distribution estimated using the across-trial mean (μ) and standard deviation (σ) of the spike count per stimulus in the 30 ms neural response window. A parameterized “population” distribution was constructed for each neuron using a Gamma distribution, Γ (N*α, Θ), where α = μ²/σ², Θ = σ²/μ, and N is the assumed number of identical neurons that share response properties similar to the experimentally recorded neuron. Results were reported using N = 10 but results were qualitatively similar for N = 1, 5, and 20. A total of 1,000 samples (trials) were drawn from the “population” distribution.

ROC curves were generated by sliding the threshold across the noise and signal distributions, generating a range of false alarm rates versus hit rates that reflect the degree of overlap of the distributions. The area under the ROC curve (AUROC) was used as a performance metric for theoretical detectability. Note that higher AUROC values indicate better performance as an AUROC value of 1 corresponds to zero overlap between signal and noise distributions (an ideal detector). An AUROC value of 0.5 corresponds to complete overlap between distributions, hence operating at chance.

For theoretical discriminability, the ability of an ideal observer to discriminate between five different velocities (50, 300, 450, 600, 1,200°/s) was quantified based on the observed S1 activity. S1 recordings of L4 RS and L4 FS neurons with at least 20 stimulus trials per velocity condition were used to generate a pool of 205 L4 RS neurons and 146 L4 FS neurons. Stimulus trials, comprising a minimum of 20 trials per condition, were amalgamated from randomly sampled cortical neurons. These datasets were employed to construct spike count distributions, which were subsequently parameterized using a gamma distribution. The resulting distributions were input into linear discriminant analysis (LDA) classifiers using MATLAB's “fitcdiscr” command to evaluate theoretical discrimination performance. This process was iterated 1,000 times for each stimulus condition. We conceptualize the ideal observer as an LDA decoder employing maximum likelihood (ML) estimation to allocate an observed response to a specific stimulus velocity. The accuracy of this classification process forms a performance matrix. Given the five stimulus velocities employed here, the performance matrix is a 5 × 5 matrix where the (i,j) element represents the probability of assigning the observed response to stimulus s_j when the actual stimulus is s_i. The theoretical discrimination performance was quantified as the sum of all diagonal elements of the performance matrix. Given that the five velocities were presented with equal probability, achieving chance discrimination corresponds to a probability of 0.2.

Statistical analyses

In order to better visualize the trends in the data, the error bars presented throughout the manuscript represent either the standard error of the mean (SEM) or confidence intervals. However, all comparisons of measures of the neuronal response across stimulus conditions were assessed with the two-sided Wilcoxon signed-rank test with Bonferroni’s correction to correct for multiple comparisons, which is independent of the visualization. Additionally, comparisons of the decay profile of individual neurons fit with single versus bi-exponential functions were assessed with the one-sided Wilcoxon signed-rank test.

Experimental design

This study consists of a within-subject design such that all comparisons are between the same set of neurons, representative of multiple animals of both sexes (see above, Animals).

Data accessibility

The processed data generated in this study have been deposited in the Zenodo database and can be accessed here: https://doi.org/10.5281/zenodo.13236981.

Previous Section Next Section

Results

Differential rapid sensory adaptation in VPm and S1

To assess the effects of adaptive sensory inputs on thalamocortical activity, we recorded extracellular neuronal spiking in the vibrissa lemniscal pathway of lightly isoflurane-anesthetized, head-fixed mice using multichannel silicon probes. We measured the activity of populations of well-isolated, single neurons in the VPm nucleus of the thalamus as well as in layer 4 of the primary somatosensory “barrel” cortex (S1 L4). S1 neurons were further parsed into putative excitatory regular spiking (RS) and putative inhibitory fast spiking (FS) neurons based on the trough-to-peak time of the spike waveform (Fig. 1A; see Materials and Methods). We assessed the neuronal responses to a 10 Hz pattern of deflections of a single whisker on the contralateral side of the face using a computer-controlled actuator for 10 s (each deflection is a transient, 300°/s sawtooth whisker deflection; see Materials and Methods).

View larger version:

Download as PowerPoint Slide

Figure 1.

Differential rapid sensory adaptation in VPm and S1. A, 32-channel silicon probe recordings were performed in the ventro-posteromedial nucleus of the thalamus (VPm) and in layer 4 (L4) of the primary somatosensory cortex (S1) in lightly isoflurane-anesthetized mice. The mean ± SEM waveforms of example well-isolated VPm, putative excitatory L4 regular spiking (RS), and putative inhibitory L4 fast spiking (FS) typical of each population are shown in green, orange, and pink, respectively. RS and FS cortical neurons were identified based on the trough to peak time of their spike waveforms. Precise deflections were delivered to single whiskers for sensory input using a computer-controlled actuator. Adaptation was induced using a train of sawtooth punctate sensory stimuli (300°/s) delivered at 10 Hz for 10 s (see Materials and Methods). B, Peristimulus time histograms (PSTHs) of the example neurons from A in response to 10 Hz repetitive sensory stimulation. C, Overlaid PSTHs of three example VPm neurons’ sensory responses to the 1st and 100th sensory stimulus in the 10 Hz repetitive stimulus train (solid and dashed lines, respectively), each with differing response profiles. The stimulus is delivered at time 0 s in each plot. D, The distribution of responsive neurons from each population that exhibited a decrease, an increase, or no change in sensory response between the 1st and 100th sensory stimulus. Change was determined using the Wilcoxon signed-rank test on the per-trials spike count in the 30 ms neural response window for the 1st versus 100th stimulus in the repetitive train. E, Grand-averaged mean ± SEM evoked response for the population of VPm, L4 RS, and L4 FS neurons to each stimulus in the 10 s repetitive stimulus train, normalized to the maximum response of the first three stimuli (see Materials and Methods). F, The mean ± SEM adaptation ratio for each neuron population after 1.5 and 10 s of repetitive stimulation, defined as the ratio of the mean spike count for the last three stimuli divided by the mean spike count for the first three stimuli. (1.5 s adapted: VPm: 0.70 ± 0.02, RS: 0.56 ± 0.02; FS: 0.65 ± 0.02; VPm vs RS: p = 2.60 × 10⁻⁷, VPm vs FS: p = 6.82 × 10⁻², RS vs FS: p = 3.96 × 10⁻⁴; 10 s adapted: VPm: 0.48 ± 0.02, RS: 0.34 ± 0.01; FS: 0.41 ± 0.02; VPm vs RS: p = 1.52 × 10⁻¹², VPm vs FS: p = 8.25 × 10⁻⁴, RS vs FS: p = 1.35 × 10⁻³. All statistical comparisons were performed using the Wilcoxon rank-sum test with Bonferroni’s correction; *** indicates p < 3.3 × 10⁻⁴, ** indicates p < 3.3 × 10⁻³, * indicates p < 1.67 × 10⁻²).

The grand-averaged peristimulus time histograms (PSTHs) in Figure 1B show the effects of sensory adaptation induced by 10 Hz repetitive stimulation on the spiking activity of example VPm, S1 L4 RS, and L4 FS neurons. As qualitatively observed for these example neurons, the sensory response of most recorded neurons in each neuronal population decreased over the timecourse of the repetitive stimulation. However, some neurons exhibit no overall change in stimulus-evoked spiking, and others an increase, as demonstrated in VPm by the overlaid PSTHs for the 1st and 100th stimuli in the adapting stimulus train in Figure 1C. All sensory-responsive VPm and S1 L4 neurons were therefore parsed into three distinct categories—increased firing, decreased firing, and no change in firing—based on the cumulative spiking evoked within 30 ms of the 1st and 100th deflection of the adapting stimulus. We found that 66, 95, and 83% of VPm, RS, and FS neurons, respectively, decreased their sensory response between the 1st and 100th deflection (Fig. 1D). Note that the results here and throughout all analyses in this study were obtained for a specific whisker stimulation frequency and velocity; the responses to other conditions are likely different in detail (see Discussion).

Despite some variability across neurons, we found that most responsive neurons across these regions responded most vigorously to the first few deflections in the stimulus train and adapted strongly to repetitive sensory stimuli. This is evident in the normalized population-averaged evoked response, calculated for each deflection in the train as the trial-averaged number of spikes in the 30 ms response window following the deflection (Fig. 1E). The sensory responses were normalized by the unadapted response, or the maximum response to the first three deflections, hereafter referred to as the control response. For each population, the response amplitude was reduced between 0 and 1.5 s, with a more gradual reduction in responsiveness from 1.5 to 10 s. Hence, throughout the rest of the study, we exclusively analyzed neurons whose sensory responses decreased between the control and the last stimulus in the 10 s train.

As further evidenced in Figure 1E, L4 RS and FS neurons adapted to a greater degree than VPm neurons in terms of their normalized response amplitude. Consistent with previous studies (Simons and Carvell, 1989; Gabernet et al., 2005; Heiss et al., 2008; Wright et al., 2021), the adaptation was even more pronounced in the RS population compared with both the VPm and the FS population when considering the adaptive decrease relative to the control sensory response. To quantify this, we calculated the adaptation ratio, or the ratio of the mean spike count in response to the last three stimuli in the repetitive stimulus train divided by the mean spike count in response to the first three stimuli. For the first 1.5 s of the adapting train, the adaptation ratios were 0.70, 0.56, and 0.65 for the VPm, RS, and FS populations, respectively. At the 10 s time point in the adapting train, the adaptation ratios decreased to 0.48, 0.34, and 0.41 (Fig. 1F). Given these observations, we sought to further investigate the adaptation dynamics in both S1 neuron subtypes and VPm thalamus at three time points of interest: the control time point as well as both the 1.5 and 10 s time points.

Multiple timescales underlie the adaptation dynamics of cortical neurons in response to repetitive sensory stimulation

Figure 2, A and B, shows the grand-averaged PSTHs for the population of L4 RS and FS neurons, respectively (note the different vertical scales for RS vs FS). Generally, both cortical subpopulations generated short-latency, transient spiking in response to the control stimulus. There was a qualitatively observable increase in latency and decrease in spiking response as a function of the duration of the adapting stimulus. Indeed, the mean first-spike latency for the L4 RS population increased from 12.2 to 19.3 ms between the Control and 1.5 s time point, with a further increase to 20.2 ms at the 10 s time point, corresponding to a 58 and 5% increase in latency between the time points. For the L4 FS population, the first-spike latency increased from 10.35 to 16.51 to 18.09 ms, corresponding to a 59 and 10% increase in latency between the time points (Fig. 2C). Similarly, the mean evoked response at the three time points of interest decreased from 0.50 to 0.22 to 0.14 spks/stim for the L4 RS population, corresponding to a 56 and 36% decrease in spiking response between the time points and from 0.72 to 0.41 to 0.30 spks/stim for the L4 FS population, corresponding to a 43 and 27% decrease in spiking response between the time points (Fig. 2D).

View larger version:

Download as PowerPoint Slide

Figure 2.

Multiple timescales underlie the adaptation dynamics of cortical neurons in response to repetitive sensory stimulation. A, Grand-averaged PSTHs of S1 layer 4 RS neurons to the 10 Hz repetitive stimulus at the three time points of interest: Control (1st stimulus in the repetitive train), 1.5 s adapted (15th stimulus), and 10 s adapted (100th stimulus; note that only sensory-responsive neurons whose sensory response significantly decreased between the Control and 10 s adapted stimuli were included). B, Same as A, but for the L4 FS population. C, Mean ± SEM first-spike latency of the L4 RS and FS populations to repetitive sensory stimuli at the three time points of interest (RS: Control: 12.21 ± 0.15 ms, 1.5 s adapted: 19.31 ± 0.19 ms, 10 s adapted: 20.22. ± 0.20 ms; Control vs 1.5 s adapted: p = 6.40 × 10⁻⁴⁰, 1.5 s adapted vs 10 s adapted: p = 6.62 × 10⁻⁷, Control vs 10 s adapted: p = 5.85 × 10⁻⁴⁰; FS: Control: 10.35 ± 0.13 ms, 1.5 s adapted: 16.51 ± 0.21 ms, 10 s adapted: 18.09 ± 0.25 ms; Control vs 1.5 s adapted: p = 1.04 × 10⁻³¹, 1.5 s adapted vs 10 s adapted: p = 5.31 × 10⁻¹⁹, Control vs 10 s adapted: p = 6.26 × 10⁻³³; all statistical comparisons were performed using the Wilcoxon signed-rank test with Bonferroni’s correction; *** indicates p < 3.3 × 10⁻⁴). D, Mean ± SEM evoked firing rate to repetitive sensory stimuli at the three time points of interest for the L4 RS and FS populations (RS: Control: 0.50 ± 0.02 spks/stim, 1.5 s adapted: 0.22 ± 0.01 spks/stim, 10 s adapted: 0.14 ± 0.01 spks/stim; Control vs 1.5 s adapted: p = 1.65 × 10⁻³⁵, 1.5 s adapted vs 10 s adapted: p = 3.36 × 10⁻³¹, Control vs 10 s adapted: p = 7.77 × 10⁻⁴⁰; FS: Control: 0.72 ± 0.03 spks/stim, 1.5 s adapted: 0.41 ± 0.0.03 spks/stim, 10 s adapted: 0.30 ± 0.02 spks/stim; Control vs 1.5 s adapted: p = 7.56 × 10⁻²⁶, 1.5 s adapted vs 10 s adapted: p = 5.48 × 10⁻²⁸, Control vs 10 s adapted: p = 7.58 × 10⁻³¹; all statistical comparisons were performed using the Wilcoxon signed-rank test with Bonferroni’s correction; *** indicates p < 3.3 × 10⁻⁴). E, The normalized mean evoked response for the L4 RS population over the timecourse of repetitive stimulation overlaid with a single, bi-exponential, and tri-exponential model fit (purple, blue, and red lines, respectively; see Materials and Methods). The evoked response is normalized per neuron by the maximum response to the first three stimuli before taking the average across all neurons. To account for differences in the number of parameters between models, goodness of fit was assessed with the Akaike information criterion (AIC); the smallest AIC value indicates the best fit. Inset, semi-log plot of the normalized evoked response. The black lines denote a linear fit over each segment of the data; the non-zero slopes suggest bi-exponential behavior (τ₁= 0.92 s, τ₂= 23.20 s, offset = 0.15). F, Same as E, but for the L4 FS population (τ₁= 0.91 s, τ₂= 29.62 s, offset = 0.12). G, The AIC values for single versus bi-exponential functions fit to the adaptation timecourse of individual L4 RS neurons (single exponential mean = −233.8; bi-exponential mean = −242.2; the p value indicates the comparison between the single and bi-exponential model fits assessed for all neurons using the one-sided Wilcoxon signed-rank test). H, Same as G, but for individual L4 FS neurons (single exponential mean = −253.8; bi-exponential mean = −264.9). I, Distributions of the model parameters for bi-exponential fits of individual L4 RS neuron timecourses. The first exponent for each neuron is the exponent with the fastest time constant. Neurons that were best fit with only one exponential are not included in the distributions for the second exponential model parameters. Note the difference in horizontal axis range across each model parameter. The triangle indicates the median across neurons for each model parameter. J. Same as I, but for individual L4 FS neurons.

Given the nonsaturating nature of the observed adaptation dynamics, we quantified the timescales of the decay in responsiveness of the cortical populations by fitting exponential models with a constant offset to the normalized population-averaged mean evoked response as a function of stimulus duration (see Materials and Methods). Although adaptation is widely characterized as a simple, single exponential decay (Chung et al., 2002; Khatri et al., 2004; Patterson et al., 2013; Kheradpezhouh et al., 2017), we found that the declining profile of the adaptation responses over the 10 s repetitive stimulation was well characterized with bi-exponential functions. Importantly, the added degrees of freedom of a single exponential function could trivially result in improved model fits. To account for these extra parameters, we computed the AIC, a relative value for which the lowest measure indicates the best fit while accounting for the number of parameters (see Materials and Methods). We found that the bi-exponential models better fit the L4 RS and FS population timecourses compared with the single exponential model (L4 RS: single exponential AIC = −516.2, bi-exponential AIC = −582.2; L4 FS: single exponential AIC = −511.1, bi-exponential AIC = −617.9). Further, the tri-exponential provided marginal improvement which, coupled with the penalty for the extra parameters, resulted in a larger AIC value and is thus a worse fit compared with the bi-exponential model (Fig. 2E,F; L4 RS tri-exponential AIC = −578.5; L4 FS tri-exponential AIC = −614.9). The bi-exponential decaying trend of the adaptation timecourse was even more evident when plotted on semi-log axes, where the plot reveals two strikingly clear linear regimes with negative slopes, suggesting two distinct timescales of exponential decay (figure insets).

Figure 2, G and H, shows the AIC values for models fit to the adaptation timecourse of individual neurons. Importantly, even at the level of individual neurons, L4 RS and FS neurons decay in a bi-exponential manner, such that the AIC values for bi-exponential model fits are significantly lower than those for the single exponential fits (RS: single exponential mean = −233.8; bi-exponential mean = −242.2; FS: single exponential mean = −253.8; bi-exponential mean = −264.9). We then assessed the bi-exponential model parameters for individual S1 neurons. We found that 72 of 233 individual L4 RS neurons (30.9%) and 60 of 190 L4 FS neurons (31.6%) were not best fit with a second exponential. However, the remaining majority of neurons evolve with a faster first time constant (median of 0.62 s for L4 RS neurons, and 0.70 s for L4 FS) and an order of magnitude slower second time constant (median of 3.92 s for L4 RS neurons, and 14.4 s for L4 FS neurons). Interestingly, we observed that the scaling coefficients of the first faster time constants are larger (median of 0.55 normalized spikes/stimulus for L4 RS, 0.52 normalized spikes/stimulus for L4 FS; normalized to the maximum response to the first three stimuli in the adapting train) compared with that of the second slower time constant (median of 0.28 normalized spikes/stimulus for L4 RS, 0.28 normalized spikes/stimulus for L4 FS). This suggests that the first faster time constant contributes more to the neurons’ absolute amount of firing rate adaptation, consistent with qualitative observations. Finally, the offset for the exponential fits was 0.19 and 0.20 for L4 RS and FS neurons, respectively (Fig. 2I,J), accounting for any residual effects not accounted for by the exponentials. It is important to note that the overall degree of adaptation reflected in the model is a combination of the offset and the residual effects of the decaying exponentials within the 10 s analysis window.

Multiple timescales underlie the adaptation dynamics of thalamic firing rate and population synchrony in response to repetitive sensory stimulation

The clear multi-timescale adaptation dynamics observed in S1 opens the possibility of multiple factors contributing to these observed cortical dynamics. A likely important contributor is the spiking patterns of the VPm neurons that provide the direct input to the cortical layer 4 neurons, in both the mean firing rate and population synchrony of this highly convergent input. Toward characterizing the adaptation dynamics in VPm, Figure 3A shows the grand-averaged PSTHs for the VPm population in response to the adapting sensory stimulus at the three time points of interest. Like S1 neurons, VPm neurons generated short-latency, transient spiking in response to the Control stimulus with shifts in latency and spiking response from the Control to the 1.5 and 10 s time points. When quantifying these shifts for the population, the mean first-spike latency increased from 9.16 to 12.88 to 14.01 ms for the three time points of interest (Fig. 3B). These shifts in latency correspond to a 40% increase from Control to the 1.5 s time point and a further increase of 8.75% between the 1.5 s the 10 s time points. Additionally, the mean evoked response decreased from 0.81 to 0.52 spks/stim between the Control and 1.5 s time point, with an additional decrease to 0.38 spks/stim from the 1.5 s to the 10 s time point (Fig. 3C), corresponding to 36 and 27% decreases in sensory response, respectively. Taken together, adaptation dynamics in the VPm also did not saturate between 1.5 and 10 s but rather continued to progress throughout the duration of the adapting stimulus.

View larger version:

Download as PowerPoint Slide

Figure 3.

Multiple timescales underlie the adaptation dynamics of thalamic firing rate in response to repetitive sensory stimulation. A, Grand-averaged PSTHs of VPm neurons to the 10 Hz repetitive stimulus at the three time points of interest: Control (1st stimulus in the repetitive train), 1.5 s adapted (15th stimulus), and 10 s adapted (100th stimulus; note that only responsive neurons whose sensory response significantly decreased between the Control and 10 s adapted stimuli were included). B, Mean ± SEM first-spike latency of the VPm population to repetitive sensory stimuli at the three time points of interest (Control: 9.16 ± 0.15 ms, 1.5 s adapted: 12.88 ± 0.20 ms, 10 s adapted: 14.01 ± 0.21 ms; Control vs 1.5 s adapted: p = 1.63 × 10⁻³¹, 1.5 s adapted vs 10 s adapted: p = 1.39 × 10⁻¹⁷, Control vs 10 s adapted: p = 1.41 × 10⁻³¹; all statistical comparisons were performed using the Wilcoxon signed-rank test with Bonferroni’s correction; *** indicates p < 3.3 × 10⁻⁴). C, Mean ± SEM evoked firing rate of the VPm population to repetitive sensory stimuli at the three time points of interest (Control: 0.81 ± 0.05 spks/stim, 1.5 s adapted: 0.52 ± 0.03 spks/stim, 10 s adapted: 0.38 ± 0.02 spks/stim. Control vs 1.5 s adapted: p = 1.15 × 10⁻¹³; 1.5 s adapted vs 10 s adapted: p = 1.44 × 10⁻²³, Control vs 10 s adapted: p = 1.43 × 10⁻²⁷, all statistical comparisons were performed using the Wilcoxon signed-rank test with Bonferroni’s correction; *** indicates p < 3.3 × 10⁻⁴). D, The normalized mean evoked response for the VPm population over the timecourse of repetitive stimulation overlaid with a single, bi-exponential, and tri-exponential model fit (purple, blue, and red lines, respectively). The evoked response is normalized per neuron by the maximum response to the first three stimuli before taking the average across all neurons. Inset, semi-log plot of the normalized evoked response. The black lines denote a linear fit over each segment of the data; the non-zero slopes suggest bi-exponential behavior (τ₁= 0.61 s, τ₂= 10.3 s, offset = 0.38). E, The AIC values for single versus bi-exponential functions fit to the adaptation timecourse of individual VPm neurons (single exponential mean = −252.2; bi-exponential mean = −263.2 the p value indicates the comparison between the single and bi-exponential model fits assessed for all neurons using the one-sided Wilcoxon signed-rank test). F, Distributions of the model parameters for bi-exponential fits of individual L4 RS neuron timecourses. The first exponential for each neuron is the exponential with the fastest time constant. Neurons that are best fit with only one exponent are not included in the distributions for the second exponential model parameters. Note the difference in horizontal axis range across each model parameter. The triangle indicates the median across neurons for each model parameter.

To characterize the timescale of the decay in responsiveness of VPm neurons, we again fit exponential models plus a constant offset to the normalized population-averaged mean evoked response as a function of stimulus duration. We found that the declining profile of the adaptation response over the 10 s repetitive stimulation was best characterized with a bi-exponential (AIC = −666.7) compared with a single exponential function (AIC = −492.3) or a tri-exponential (AIC = −662.7; Fig. 3D). Again, the bi-exponential nature of the decay became strikingly clear when plotting the normalized evoked responses on semi-log axes, highlighting two linear regions (figure insets).

We next assessed the adaptation timecourse of individual VPm neurons. Figure 3E shows the AIC values measured for fits of the adaptation timecourse of individual VPm neurons. While some of the VPm neurons are equally well fit by single and bi-exponential models, the mean AIC value is significantly lower for the bi-exponential models (single exponential mean = −252.2; bi-exponential mean = −263.2) indicating that, overall, the population of individual VPm neurons is better fit with bi-exponential functions than single exponential functions even at the level of individual neurons (Fig. 3E). We next assessed the model parameters for the adaptation timecourse of individual neurons. We found that 64 of 184 VPm neurons (34.8%) were not best fit with a second exponential. However, the remaining majority of neurons evolve with a faster first time constant (median of 0.55 s) and an order of magnitude slower second time constant (median of 12.14 s). Similar to S1 neurons, the first exponent with the faster time constant contributed more to the absolute amount of firing rate adaptation, with a median scaling coefficient of 0.47 normalized spikes/stimulus across all VPm neurons compared with a coefficient of 0.31 normalized spikes/stimulus for the second and slower exponential decay (normalized to the maximum response to the first three stimuli in the adapting train). Additionally, the median offset was 0.2 normalized spikes/stimulus (Fig. 3F).

A number of studies have shown that S1 can be highly sensitive to changes in the synchrony of the convergent thalamic inputs (Pinto et al., 2000; Swadlow and Gusev, 2001; Bruno and Sakmann, 2006; Wang et al., 2010a,b; Ollerenshaw et al., 2014). We thus sought to also characterize the adaptation profile of VPm synchrony. We quantified the synchrony between every pair of simultaneously recorded neurons having maximal responsiveness to the same whisker (i.e., likely in the same thalamic barreloid). Synchrony was computed by generating a cross-correlogram (CCG) and measuring the occurrence of coincident spikes within a ±7.5 ms window, normalized by the number of spikes used in the analysis (see Materials and Methods; Fig. 4A). This normalization accounts for the overall firing rate which is important given that synchrony and firing rate naturally covary. Note that due to the sensitivity of the synchrony measurements to datalength, synchrony measurements here are presented only for the population mean. First examining changes in synchrony in response to the adapting stimuli at the three time points of interest, the VPm population synchrony decreased from 0.89 to 0.81 to 0.76 (unitless quantity; see Materials and Methods; Fig. 4B). Further, as with the mean firing rate, the synchrony evolved in a bi-exponential manner (AIC = −707.9) and was not better fit with a single exponential (AIC = −637.7) or a tri-exponential (AIC = −704.1; Fig. 4C).

View larger version:

Download as PowerPoint Slide

Figure 4.

Multiple timescales underlie the adaptation dynamics of thalamic synchrony in response to repetitive sensory stimulation. A, VPm population synchrony is measured by generating pairwise cross-correlograms (CCGs) and quantifying the area in the ±7.5 ms window normalized by the area under the entire CCG (see Methods). B, Mean ± SEM synchrony of the VPm population to repetitive stimuli at the three time points of interest (Control: 0.89 ± 4.01 × 10⁻³, 1.5 s adapted: 0.81 ± 4.30 × 10⁻³, 10 s adapted: 0.76 ± 5.52 × 10⁻³. Control vs 1.5 s adapted: p = 6.90 × 10⁻⁹²; 1.5 s adapted vs 10 s adapted: p = 1.52 × 10⁻²⁰, Control vs 10 s adapted: p = 3.90 × 10⁻¹⁰⁷, all statistical comparisons were performed using the Wilcoxon signed-rank test with Bonferroni’s correction; *** indicates p <3.3 × 10⁻⁴). C, The normalized mean evoked synchrony for the VPm population over the timecourse of repetitive stimulation overlaid with a single and a bi-exponential model fit (purple and blue dashed lines, respectively). The synchrony is normalized per pair of neurons by the maximum synchrony in response to the first three stimuli before taking the average across all pairs of neurons. Inset, semi-log plot of the normalized population synchrony. The black lines denote the linear fit over each segment of the data; the non-zero slopes suggest bi-exponential behavior (τ₁= 0.19 s, τ₂= 5.22 s, offset = 0.75 normalized spks/stim).

Multiple timescales underlie the adaptation dynamics of thalamocortical synaptic efficacy in response to repetitive stimulation

The above analyses suggest a relationship between cortical L4 neurons and the VPm neurons from which the L4 neurons receive direct sensory input but do not measure this relationship explicitly. To evaluate the dynamics of the transformation across the thalamocortical synapse, we next examined the effects of the adapting stimulus on synaptic efficacy of putatively connected S1 pairs recorded simultaneously using previously established methods for identifying monosynaptic connectivity in vivo (Liew et al., 2021). The left panel of Figure 5A shows raster plots of example topographically aligned VPm, L4 RS, and L4 FS neurons that were simultaneously recorded in response to a weak, sinusoidal input stimulus. Note that the overall firing rate of all the neurons was elevated but not generally stimulus-locked; this is critical for identifying putatively connected pairs. The spiking evoked by this weak sinusoidal stimulus was used to generate raw CCGs between every possible pair of topographically aligned VPm→L4 RS neurons. Shuffled CCGs were generated by randomizing the trials of the VPM and S1 spiking activity relative to each other, effectively correcting for any stimulus-induced elements of the CCG. Finally, shuffle-corrected (SC) CCGs were generated by subtracting the shuffled CCGs from the raw CCGs (see Materials and Methods). As is the case for the SC CCGs of the example pairs of neurons shown in the right panel of Figure 5A, a narrow peak within the 1–3 ms window with prominence greater than 3.5 standard deviations of the shuffled CCG suggests monosynaptic connectivity (see Materials and Methods; Liew et al., 2021). Of the 1,199 recorded VPm→RS thalamocortical pairs that were found to be topographically aligned, 297 were inferred to be monosynaptically connected (24.8%). Similarly, of the 681 recorded VPm→FS thalamocortical pairs that were found to be topographically aligned, 319 were inferred to be monosynaptically connected (46.8%). Subsequent analyses were restricted to these putatively connected thalamocortical neuron pairs.

View larger version:

Download as PowerPoint Slide

Figure 5.

Multiple timescales underlie the adaptation dynamics of thalamocortical synaptic efficacy in response to repetitive sensory stimulation. A, Left, Spike raster plots of a simultaneously recorded VPm, L4 RS, and L4 FS neuron in response to a weak sinusoidal whisker stimulus (4 Hz, mean velocity: 25−50°/s) in order to elevate spiking activity without driving correlated, stimulus-locked spiking across neurons. Right, The shuffle-corrected (SC) cross-correlograms (CCGs) for the L4 RS (top) and FS (bottom) neuron, each relative to the VPm neuron's spiking. The SC CCGs are generated by subtracting the shuffled CCGs from the raw CCGs. Raw CCGs are generated using the spikes elicited by the sinusoidal whisker stimulus and the shuffled CCGs are generated by randomizing the trials of the VPM and S1 spiking activity relative to each other. The gray lines denote 3.5 standard deviations of the shuffled distribution. A peak in the SC CCGs in the 1–3 ms window (window highlighted in light brown) indicates that these pairs of neurons are putative monosynaptically connected (Liew et al., 2021). B, Spike rasters of the putative monosynaptically connected neurons from A in response to 10 Hz repetitive sensory stimulus (300°/s). C, The grand-averaged Control (left), 1.5 s adapted (middle), and 10 s adapted (right) SC (black) and raw (purple) CCGs for all putative monosynaptically connected VPm→L4 RS neuron pairs. The 1–4 ms window used in subsequent analysis is highlighted in light brown. D, Same as C, but for all putative monosynaptically connected VPm→L4 FS pairs. E, Mean ± SEM synaptic efficacy for all putative monosynaptically connected VPm→L4 RS pairs to repetitive sensory stimuli at the three time points of interest. Given the noisy nature of synaptic efficacy measurements, the control datapoint represents the average efficacy across first the stimuli, the 1.5 s adapted datapoint represents the average efficacy across the 14th–16th stimuli (1.4–1.6 s), and the 10 s adapted timepoint represents the average efficacy across the 98th–100th stimuli (9.8–10 s; Control: 2.1 × 10⁻² ± 1.7 × 10⁻³, 1.5 s adapted: 8.6 × 10⁻³ ± 9.4 × 10⁻⁴, 10 s adapted: 6.4 × 10⁻³ ± 8.1 × 10⁻⁴. Control vs 1.5 s adapted: p = 1.2 × 10⁻¹¹; 1.5 s adapted vs 10 s adapted: p = 1.2 × 10⁻³, Control vs 10 s adapted: p = 2.2 × 10⁻¹⁵, all comparisons were performed using the Wilcoxon signed-rank test with Bonferroni’s correction; *** indicates p < 3.3 × 10⁻⁴, ** indicates p < 3.3 × 10⁻³). Synaptic efficacy is defined for each adaptation time point as the number of synchronous spikes within the 1−3 ms window divided by the number of VPm spikes contributing to the analysis (see Materials and Methods). F, Same as E, but for all putative monosynaptically connected VPm→L4 FS pairs (Control: 4.1 × 10⁻² ± 2.2 × 10⁻³, 1.5 s adapted: 3.2 × 10⁻² ± 1.9 × 10⁻³, 10 s adapted: 2.7 × 10⁻² ± 1.9 × 10⁻³. Control vs 1.5 s adapted: p = 4.3 × 10⁻⁵; 1.5 s adapted vs 10 s adapted: p = 3.8 × 10⁻², Control vs 10 s adapted: p = 1.6 × 10⁻⁸, all comparisons were performed using the Wilcoxon signed-rank test with Bonferroni’s correction; *** indicates p < 3.3 × 10⁻⁴). G, The normalized mean synaptic efficacy for all putative monosynaptically connected VPm→L4 RS pairs over the timecourse of repetitive sensory stimulation overlaid with a single and a bi-exponential model fit (blue and purple dashed lines, respectively). The efficacy is normalized per neuron by the maximum response to the first three stimuli before taking the average across all neurons. Inset, semi-log plot of the normalized population synaptic efficacy. The purple and blue lines denote the linear fit over each segment of the data; the non-zero slopes suggest bi-exponential behavior (τ₁= 0.16 s, τ₂= 2.1 s, offset = 0.25). H, Same as G, but for all putative monosynaptically connected VPm→L4 FS pairs (τ₁= 0.19 s, τ₂= 12.4 s, offset = 0.44 normalized spks/stim).

Responses to the 10 Hz adapting stimulus for the same example VPm, L4 RS, and L4 FS neurons in Figure 5A are shown in Figure 5B. For each of the three time points of interest, we assessed the synaptic efficacy of each putatively connected thalamocortical pair, defined as the probability of an S1 neuron spike following a spike of the presynaptic VPm neuron. More specifically, using the spiking activity in a 80 ms window following each of the time points of interest in the repetitive stimulus, synaptic efficacy was calculated as the number of synchronous spikes in the SC CCG within the 1–3 ms window following each VPm spike divided by the total number of spikes in the SC CCG (Swadlow and Gusev, 2001).

Figure 5, C and D, shows a clear drop in the peak of the grand-averaged SC CCG generated from the spiking responses at the three time points of interest in the 10 Hz repetitive stimulation for VPm→L4 RS and VPm→L4 FS putatively connected pairs, respectively. Indeed, synaptic efficacy decreased across the three time points for both VPm→L4 RS and VPm→L4 FS, summarized in Figure 5E and F. Note that the apparent decrease in synaptic efficacy from 1.5 to 10 s for the VPm→L4 FS pairs was not statistically significant. When plotting the synaptic efficacy continuously computed over the timecourse of stimulation, we observe a decaying trend, with an early rapid drop of 81 and 37% at the 1.5 s time point compared with control for the VPm→L4 RS and VPm→L4 FS pairs, respectively. Following this early drop, synaptic efficacy gradually declined during the latter part of stimulation, corresponding to a 50 and 5% drop at the 10 s time point compared with the 1.5 s time point for the VPm→L4 RS and VPm→L4 FS pairs, respectively. Here again, through modeling we observe two distinct timescales (L4 RS: 1-exp: AIC = −287.4; 2-exp: AIC = −297.4; 3-exp: AIC = −293.5. L4 FS: 1-exp: AIC = −285.7; 2-exp: AIC = −302.9; 3-exp: AIC = −298.9), as evident in the semi-log plot insets (Fig. 5G,H). It is important to note that, like the synchrony measures of Figure 4, the synaptic efficacy measures are strongly datalength dependent and highly variable, and thus analyses were conducted for population means only. It also is important to note that since the above measures of synaptic efficacy were obtained during sensory inputs that synchronize the thalamic population, the efficacy measures could be influenced by the overall drive and synchronization of the thalamic input, and the resultant depression at the thalamocortical synapse, as well as the frequency and velocity of the whisker stimulation (see Discussion).

Direct optogenetic drive of VPm thalamus eliminates longer timescale adaptation in VPm

In our next set of experiments, we sought to explore the relationship across the thalamocortical junction through a more direct manipulation of VPm itself. Specifically, we optogenetically stimulated VPm neurons in NR133 × Ai32 mice which enables selective expression of Channelrhodopsin (ChR2) in VPm neurons (Fig. 6A). Using an optrode lowered into VPm, we presented pulsatile optogenetic stimulation of identical frequency and duration to the sensory stimulation: 10 Hz for 10 s. The LED pulse waveform was designed to elicit a response that mimics the control VPm sensory response (Fig. 6B; see Materials and Methods).

View larger version:

Download as PowerPoint Slide

Figure 6.

Direct optogenetic drive of VPm thalamus eliminates longer timescale adaptation in VPm. A, ChR2-expressing VPm neurons in lightly isoflurane-anesthetized NR133 × Ai32 mice were activated with 10 Hz repetitive LED pulses to selectively control properties of the VPm inputs while recording VPm and cortical L4 activity using 32-channel silicon probe. The histological section shows expression of ChR2 tagged with eYFP (green) in the VPm and the colocalized electrode track fluorescently labeled with DiI (red) among DAPI labelled neurons (blue). B, The mean ± SEM firing rate of the VPm population to control sensory and optostimulation (Sensory: 0.6 ± 5.6 × 10⁻² spks/stim, Opto: 0.6 ± 0.8.2e × 10⁻² spks/stim, p = 0.37, the comparison was performed using the Wilcoxon signed-rank test). C, Grand-averaged PSTHs for VPm neurons to 10 Hz opto stimulation of VPm at the three time points of interest. D, Mean ± SEM evoked response of the VPm population to repetitive optogenetic stimuli at the three time points of interest (Control: 0.62 ± 0.09 spks/stim, 1.5 s adapted: 0.44 ± 0.06 spks/stim, 10 s adapted: 0.45 ± 0.06 spks/stim. Control vs 1.5 s adapted: p = 1.72 × 10⁻⁴; 1.5 s adapted vs 10 s adapted: p = 7.17 × 10⁻¹, Control vs 10 s adapted: p = 2.50 × 10⁻⁴, all statistical comparisons were performed using the Wilcoxon signed-rank test with Bonferroni’s correction; *** indicates p < 3.3 × 10⁻⁴). E, Population-averaged evoked response for the VPm population over the timecourse of repetitive optogenetic stimulation overlaid with a single exponential model fit. Inset, semi-log plot of the normalized evoked response; the black lines denote the linear fit over one timescale plus an offset, suggesting single exponential behavior (τ₁= 0.36 s, offset = 0.61). F, Mean ± SEM synchrony of the VPm population to repetitive optogenetic stimuli at the three time points of interest (Control: 0.87 ± 1.80 × 10⁻³, 1.5 s adapted: 0.79 ± 1.88 × 10⁻³, 10 s adapted: 0.79 ± 1.76 × 10⁻³. Control vs 1.5 s adapted: p = 1.45 × 10⁻⁷; 1.5 s adapted vs 10 s adapted: p = 3.48 × 10⁻¹, Control vs 10 s adapted: p = 8.08 × 10⁻⁸, all statistical comparisons were performed using the Wilcoxon signed-rank test with Bonferroni’s correction; *** indicates p < 3.3 × 10⁻⁴). G, Population-averaged synchrony of the VPm population over the timecourse of repetitive optogenetic stimulation overlaid with a single exponential model fit. Inset, semi-log plot of the normalized evoked response; the black lines denote the linear fit over each segment of the data; the non-zero slopes suggest single exponential behavior (τ₁= 0.57 s, offset = 0.61 normalized spks/stim).

As shown by the grand-averaged PSTHs in Figure 6C, repetitive optogenetic stimulation produced a decrease in response in VPm neurons between the control and 1.5 s time point in the repetitive optogenetic stimulation. Although the optogenetic input induced an initial facilitation in the thalamic response not generally observed in the sensory adaptation, likely due to synchronization effects of the optogenetic light pulse, beyond the first 200–300 ms, the adaptation induced by the optogenetic train was similar to that observed for adaptation with the sensory stimulus (Fig. 3A). However, in contrast to the sensory adaptation, qualitatively there was a less clear difference in response between the 1.5 and 10 s time points. In response to the optogenetic stimulation at the three time points of interest, the evoked VPm population firing rate decreased from the Control to the 1.5 s condition but did not further decrease between the 1.5 and 10 s conditions (Fig. 6D). We characterized the timescale of the decay in responsiveness of VPm neurons by again fitting with exponential models the normalized population-averaged mean evoked optogenetic response as a function of stimulus duration. This time, we found that the declining profile was better characterized with a single exponential (1-exp: AIC = −479.6; 2-exp: AIC = −475.6). The single exponential nature of the decay becomes strikingly clear when plotting the normalized evoked responses on semi-log axes, highlighting only one linear region with a slope <0, suggesting a single exponential with a constant offset (Fig. 6E). It is important to note that the initial quick transient facilitation in the VPm response that was unique to the optogenetic input did not influence this quantification.

Similarly, in response to repetitive optogenetic stimulation, VPm synchrony decreased from Control to the 1.5 s adapted conditions (also similar to that for sensory adaptation—Fig. 4B), but no differences were detected between the 1.5 and 10 s adapted conditions (Fig. 6F). In addition, the timecourse of VPm synchrony to repetitive optogenetic stimulation was well fit with a single exponential and not improved with a second exponential, further evidenced by the semi-log insets (Fig. 6G; 1-exp: AIC = −497.3; 2-exp: AIC = −494.2).

Optogenetic drive of VPm results in a loss of longer timescale adaptation in S1

Having designed a condition that eliminates the longer timescale adaptation in the thalamus such that VPm synchrony and firing rate rapidly decayed before stabilizing for the remainder of the repetitive stimulus, we then considered the dynamics in S1. Figure 7, A and B, shows the grand-averaged cortical responses to optogenetic stimulation of VPm. The L4 RS and FS neurons adapted strongly to the repetitive optogenetic stimuli. At the three time points of interest, both the L4 RS and FS neuron populations’ firing rate decreased from the Control to the 1.5 s time point but did not further decrease from the 1.5 to 10 s time point (Fig. 7C,D, respectively). As evident in the normalized population-averaged mean evoked response in both cortical populations, the reduction in cortical response over the 10 s timecourse was well characterized by a single timescale, with marginal gain for a second exponential (Fig. 7E,F; 1-exp: AIC = −325.8 for L4 RS and AIC = −307.0 for L4 FS; 2-exp: AIC = −321.8 for L4 RS and AIC = −303.0 for L4 FS). When visualizing the timecourse on semi-log axes for both L4 RS and FS populations, a distinct pattern emerges: a single linear region is followed by a relatively flat line. This observation indicates that the decay in evoked responses occurred within a single timescale.

View larger version:

Download as PowerPoint Slide

Figure 7.

Optogenetic drive of VPm results in a single timecourse of adaptation in S1. A, Grand-averaged PSTHs of L4 RS neurons in response to 10 Hz repetitive optogenetic stimulation of VPm at the three timepoints of interest. B, Same as A but for L4 FS neurons. C, Mean ± SEM evoked response of the L4 RS population to repetitive optogenetic stimuli of VPm at the three time points of interest (Control: 1.07 × 10⁻¹ ± 9.30 × 10⁻³, 1.5 s adapted: 6.16 × 10⁻² ± 4.52 × 10⁻³, 10 s adapted: 5.59 × 10⁻² ± 3.49 × 10⁻³. Control vs 1.5 s adapted: p = 3.89 × 10⁻¹³; 1.5 s adapted vs 10 s adapted: p = 2.63 × 10⁻¹, Control vs 10 s adapted: p = 1.76 × 10⁻¹⁵, all statistical comparisons were performed using the Wilcoxon signed-rank test with Bonferroni’s correction; *** indicates p < 3.3 × 10⁻⁴). D, Same as C, but for the L4 FS population (Control: 2.09 × 10⁻¹ ± 2.33 × 10⁻², 1.5 s adapted: 9.46 × 10⁻² ± 9.58 × 10⁻³, 10 s adapted: 8.87 × 10⁻² ± 8.84 × 10⁻³. Control vs 1.5 s adapted: p = 3.64 × 10⁻⁹; 1.5 s adapted vs 10 s adapted: p = 3.99 × 10⁻¹, Control vs 10 s adapted: p = 2.87 × 10⁻¹¹, all statistical comparisons were performed using the Wilcoxon signed-rank test with Bonferroni’s correction; *** indicates p < 3.3 × 10⁻⁴). E, Population-averaged evoked response for the L4 RS population over the timecourse of repetitive optogenetic stimulation of VPm overlaid with a single exponential model fit. Inset, semi-log plot of the normalized evoked response; the black lines denote the linear fit over each segment of the data; the non-zero slopes suggest single exponential behavior (τ₁= 0.46 s, offset = 0.40). F, Same as E, but for the L4 FS population (τ₁= 0.47 s, offset = 0.34 normalized spks/stim).

Through these control experiments, we find that optogenetic drive of VPm that eliminates the longer timescale adaptation in both firing rate and synchrony in the VPm neuron population also eliminates the longer timescale adaptation in S1 layer 4 neurons. This suggests that the previous observations of the longer timescale adaptation in S1 neurons during sensory drive are inherited from VPm thalamus, consistent with a growing notion of VPm as the source of much of the adaptation observed in S1 (Wright et al., 2021).

Adaptation over multiple timescales degrades stimulus detection for an ideal observer of cortical activity

Having explored thalamocortical activity across multiple timescales of sensory adaptation, we next sought to investigate the functional consequences of these timescales. In previous studies of the vibrissa pathway, we have shown that repetitive stimulation lasting as brief as 1–2 s is sufficient to not only reduce the overall responsiveness of the cortical population but shift the neuronal responses toward an adaptive trade-off that favors sensory discrimination over detection (Wang et al., 2010b; Ollerenshaw et al., 2014). To determine if this holds true for longer timescales of adaptation, we probed cortical responses either in the presence or absence of a preceding adapting stimulus and leveraged signal detection theory to ascertain the ability of an ideal observer to detect a signal (stimulus present) versus noise (stimulus absent; Fig. 8A). Although the results in Figure 2 regarding the decrease in the magnitude of the sensory evoked cortical response would suggest a corresponding decrease in detectability, this is ultimately determined by both the magnitude of the response and the corresponding variability, relative to ongoing cortical activity. To directly test this, the cortical response to a 300°/s punctate stimulus and the corresponding detectability were evaluated in isolation (Control) and in the presence of preceding 10 Hz adapting stimuli of 1.5 or 10 s durations.

View larger version:

Download as PowerPoint Slide

Figure 8.

Adaptation degrades stimulus detection for an ideal observer of cortical activity. A, Cortical responses to sensory inputs of 300°/s were probed either in the presence or absence of a preceding adapting stimulus. B, Parametric signal (stimulus present) and noise (stimulus absent) population spike probability distributions for L4 RS neurons when the probe stimulus is preceded by the Control (top), 1.5 s adapted (middle), or 10 s adapted (bottom) stimulus. Parametric fits were generated using a gamma distribution on the L4 RS population response (see Materials and Methods). C, Same as in B, but for the L4 FS neuron population. D, Receiver operator characteristic (ROC) curves of the sensory response of an example L4 RS neuron when the stimulus is preceded by the Control, 1.5 s adapting, or 10 s adapting stimuli. E, Same as D, but for the L4 FS population. F, Mean ± SEM performance of an ideal observer of L4 RS activity in detecting the signal versus noise in conditions preceded by Control, 1.5 s adapting, and 10 s adapting stimuli. The performance is quantified by the area under the receiver operator characteristic curve (AUROC) performed across neurons; AUROC for control: 0.96 ± 6.36 × 10⁻³, 1.5 s adapted: 0.81 ± 1.31 × 10⁻², 10 s adapted: 0.74 ± 1.18 × 10⁻²; Control vs 1.5 s adapted: p = 5.05 × 10⁻²³, 1.5 s vs 10 s: p = 5.31 × 10⁻⁷, Control vs 10 s adapted: p = 5.68 × 10⁻²⁹; all statistical comparisons were performed using the Wilcoxon signed-rank test with Bonferroni’s correction; *** indicates p < 3.3 × 10⁻⁴). G, Same as F but for L4 FS neurons (AUROC for control: 0.96 ± 8.04 × 10⁻³, 1.5 s adapted: 0.88 ± 1.45 × 10⁻², 10 s adapted: 0.81 ± 1.63 × 10⁻²; Control vs 1.5 s adapted: p = 4.76 × 10⁻¹⁰, 1.5 s vs 10 s: p = 4.76 × 10⁻⁹, Control vs 10 s adapted: p = 1.37 × 10⁻¹⁷), all statistical comparisons were performed using the Wilcoxon signed-rank test with Bonferroni’s correction; *** indicates p < 3.3 × 10⁻⁴).

To quantify cortical sensory responses under the signal detection framework, we constructed parameterized probability distributions of the signal and noise spike counts resulting from the probe stimulus preceded by the Control and adapting stimulus conditions. The signal distributions were estimated by drawing samples from gamma distribution fits of the observed cortical spike counts within a 30 ms poststimulus response window for a 300°/s stimulus. The noise distributions were estimated from baseline activity 30 ms prior to stimulus onset (see Materials and Methods). We find that adaptation pushed the signal distribution of the evoked response closer to the noise distribution at each time point of interest for both L4 RS and FS populations (Fig. 8B,C, respectively).

To predict the potential impact of adaptation on stimulus detectability, we quantified the extent to which an ideal observer of cortical spiking activity accurately detects the presence of a signal from background noise when the signal is preceded by the Control and adapted conditions. Thus, we applied a sliding threshold on the parametric distributions to generate a receiver operator characteristics (ROC) curve which shows the hit rate versus the false alarm rate for an ideal observer of the neuronal activity. For ROC curves, a perfect diagonal indicates that the ideal observer is operating at chance, and deviations above the diagonal indicate improved performance. Figure 8, D and E, shows the ROC curves for an example L4 RS and FS neuron, respectively, for each of the three stimulus conditions. For both populations, adapting stimuli preceding the probe stimulus brought the ROC curve closer to the diagonal, reflective of a loss in theoretical detection performance. This loss was greatest when the probe stimulus was preceded by 10 s of an adapting stimulus.

To quantify this loss in theoretical detection performance, we measured the area under the ROC curve (AUROC) for each neuron. An AUROC value of 1 corresponds to zero overlap between signal and noise distribution (an ideal detector). An AUROC value of 0.5 corresponds to complete overlap between distributions, hence the ideal observer operating at chance. We found that the detectability of both RS and FS neurons significantly decreased when preceded by Control versus 1.5 s adapting stimuli and further decreased when preceded by 10 s adapting stimuli (Fig. 8F,G). Consistent with the observation in Figure 1F that RS neurons had lower adaptation ratios than FS neurons, we found that the theoretical detectability was generally lower for RS neurons as compared with FS neurons.

Adaptation over multiple timescales enhances stimulus discrimination for an ideal observer of cortical activity

We next evaluated the functional consequences of adaptation on stimulus discrimination across multiple timescales by presenting probe stimuli of varying velocities (50, 300, 450, 600, or 1,200°/s), preceded by either Control, 1.5 s, or 10 s adapting stimuli. First, in terms of the impact of adaptation on the velocity sensitivity properties in the cortex, adaptation attenuated the response magnitude across all velocities for cortical populations, although differently for different velocities. The velocity sensitivity was further reduced for both the RS and FS populations when the probe stimuli were preceded by the 10 s adapted stimulus compared with the 1.5 s adapted stimulus (Fig. 9A,B).

View larger version:

Download as PowerPoint Slide

Figure 9.

Adaptation enhances stimulus discrimination for an ideal observer of cortical activity. A, Stimulus response curves of the L4 RS population response to probe stimuli of varying velocities preceded by Control, 1.5 s adapting, or 10 s adapting stimuli. B, Same as A, but for the L4 FS population. C, Parametric distribution of L4 RS sensory responses to probe stimuli of varying velocities when the probe stimulus is preceded by the Control (top), 1.5 s adapting (middle), or 10 s adapting (bottom) stimuli. Parametric fits are generated using a gamma distribution of the RS population response (see Materials and Methods). D, Same as C but for the L4 FS population. E, The performance of an ideal observer of L4 RS activity in discriminating between sensory stimuli of varying velocities in conditions preceded by Control, 1.5 s adapting, and 10 s adapting stimuli (mean ±95% confidence intervals from re-sampling neurons with replacement; see Materials and Methods). The performance is quantified by the accuracy of an LDA classifier in discriminating between all velocities (Control: 0.33 ± 3.1 × 10⁻³, 1.5 s adapted: 0.35 ± 4.7 × 10⁻³, 10 s adapted: 0.36 ± 4.3 × 10⁻³. Control vs 1.5 s adapted: p = 1.7 × 10⁻², 1.5 s vs 10 s, p = 1.7 × 10⁻², Control vs 10 s adapted: p = 1.7 × 10⁻²; all statistical comparisons of confidence intervals were performed with Bonferroni’s correction). F, Same as E but for the FS population (mean ±99.9% confidence intervals from re-sampling neurons with replacement; Control: 0.34 ± 5.7 × 10⁻³, 1.5 s adapted: 0.41 ± 7.3 × 10⁻³, 10 s adapted: 0.43 ± 6.9 × 10⁻³; Control vs 1.5 s adapted: p = 3.3 × 10⁻⁴, 1.5 s vs 10 s, p = 3.3 × 10⁻⁴, Control vs 10 s adapted: p = 3.3 × 10⁻⁴).

Taking the perspective of an ideal observer, we examined the extent to which adaptation over multiple timescales affects discrimination between whisker deflections at different velocities. Shown in Figure 9C and D are parametric fits of the experimentally observed spike count distributions for the RS and FS populations, respectively. Qualitatively, the populations’ spike counts were relatively high in the Control conditions, but with highly overlapped and unordered distributions. However, while increasing prestimulus adaptation attenuated the population spikes counts, the distributions became more separated and ordered (Fig. 9C,D).

To quantify this effect, we compared the performance of the ideal observer in discriminating between different deflection velocities when preceded by Control, 1.5 s, and 10 s adapting stimuli. For each stimulus condition, we randomly selected 25 neurons from each L4 RS and L4 FS population and combined stimulus trials to form sensory response distributions of the cortical responses to various velocities (parameterized gamma distributions, see Materials and Methods). To assess discrimination performance, we implemented a holdout validation approach, reserving data from 20% of trials for testing while using the remaining 80% of trials for training. Linear discriminant analysis (LDA) classifiers were then applied to infer the most likely velocity presented based on the observed cortical response, and discrimination performance was evaluated by the accuracy of the ideal observer in making correct stimulus classifications (see Materials and Methods). This process was iterated 1,000 times to comprehensively assess discrimination performance for each stimulus condition (Control, 1.5 s, and 10 s adapted) across various combinations of neurons and trials. Given that all five velocities were presented with equal probability, a value of 0.2 reflects the ideal observer operating at chance. We found that increasing adaptation significantly improved overall discrimination performance in both RS and FS populations (Fig. 9E,F), across the early and later stages of the adaptation. Taken together, our results suggest that adaptation over multiple timescales continues to facilitate better discrimination performance by reducing the overlap between stimulus distributions by trading off the overall sensitivity to repetitive stimulation.

Previous Section Next Section

Discussion

In this study, we discovered two clear and distinct timescales of rapid adaptation across the thalamocortical circuit over the course of seconds. Several other studies have examined the properties of adaptation over a range of timepoints using various durations of repetitive stimulation (Dragoi et al., 2000; Patterson et al., 2013; Latimer et al., 2019), white noise stimulation (Fairhall et al., 2000; Lundstrom et al., 2010), and modeling approaches (Latimer and Fairhall, 2020). Here, the ability to record simultaneously from the thalamic VPm population as well as both the cortical L4 RS and FS populations allowed us to precisely examine transformations of adaptation across the thalamocortical junction through the spiking response amplitudes, thalamic population synchrony, and synaptic efficacy.

Numerous studies have identified potential circuit mechanisms that could underlie the differential adaptation from the thalamus to cortex. Potential drivers of the observed rapid cortical sensory adaptation include depression of thalamocortical (TC) synapses (Chung et al., 2002) and/or synaptic depression/facilitation of excitatory/inhibitory intracortical (IC) projections (Beierlein and Connors, 2002; Reig et al., 2006). There are also neuronal population mechanisms such as thalamic firing rate and population synchrony. All of these mechanisms may collectively shape the IC balance of excitation and inhibition. One prominent view is that short-term depression of thalamocortical synapses plays a dominant role in this type of cortical adaptation (Chung et al., 2002; Castro-Alamancos, 2004b; Katz et al., 2006). However, the rapid synaptic depression of the TC synapse reported in Chung et al. (2002) does not rule out the contribution of thalamic population synchrony observed here in shaping S1 adaptation, which was not measured in the relevant studies. Further, subsequent studies suggest that thalamocortical desynchronization plays perhaps an even more prominent role than short-term changes in the TC or IC synapses (Wright et al., 2021). In fact, many lines of evidence have demonstrated the profound effect that thalamic spike timing and synchrony has on S1 processing either under anesthesia (Pinto et al., 2000; Wang et al., 2010b; Whitmire and Stanley, 2016) or wakefulness (Wright et al., 2021; Borden et al., 2022). The optogenetic control experiments here do not rule out the possibility of TC synaptic depression playing a role in the fast timescale adaptation, but previous findings (Wright et al., 2021) suggest that this role may be overshadowed by the adaptive reductions in thalamic firing rate and synchrony that are also clearly present in the adaptation at fast timescales (0–1.5 s). What the optogenetic control experiments here do suggest is that the monotonic slow desynchronization of the thalamic population beyond the first few seconds (2–10 s) seems to play a dominant role in modulating cortical adaptation at longer timescales, with little or no influence from TC synaptic depression on this timescale. These results together suggest the possibility of a complex interaction between distinct mechanisms and that multiple mechanisms are likely coactivated over distinct periods of time to contribute to the cellular dynamics that we observed.

Differential adaptation of the putative excitatory and inhibitory cortical neurons observed here likely shapes overall cortical excitability and the integration properties of ascending sensory signals from the thalamus. Importantly, this effect has been shown during wakefulness, but the extent of the differential adaptation is strongly shaped by the nature of the adapting stimulus itself (Wright et al., 2021). The more profound depression of intracortical inhibitory synapses as compared with excitatory synapses during persistent stimulation has been shown to skew the E–I balance toward excitation (Gabernet et al., 2005; Heiss et al., 2008). This reduction in feedforward inhibition was shown to expand the cortical window of integration and affect how inputs are summed across multiple timescales of adaptation. This mechanism could potentially regulate the effects of the thalamic desynchronization observed here, offsetting the loss of thalamic synchrony through a loss in cortical sensitivity to thalamic synchrony. However, in previous work, we have shown that the desynchronization observed in the thalamus during adaptation outweighs the changes in cortical integration (Wang et al., 2010b).

This study highlights the important contributions of the rapid adaptation of thalamus in shaping cortical adaptation. What is the origin of the adaptative properties observed in sensory thalamus? For the fastest timescale component of the VPm adaptation, given the high firing rates of neurons in the brainstem (Ahissar et al., 2000), it is likely that depression of the synaptic inputs from the brainstem to thalamus play an important role in the rapid decrease in thalamic firing rate, as well as the desynchronization in the thalamic population at this timescale. One important finding here was that the longer timescale adaptation observed in S1 appears to be largely inherited from VPm. When VPm thalamus was optogenetically driven, this eliminated the slow decrease in thalamic firing rate and synchronization (Fig. 6), which subsequently eliminated the slower, later phase of adaptation in S1 (Fig. 7). One distinct possibility is that the direct optogenetic drive of VPm overcomes intrinsic properties of the thalamic neurons that are responsible for the slow drift. In brain slice experiments, thalamic neurons have been shown to exhibit a slow after-hyperpolarization that persists for several seconds after controlled depolarization (Goaillard and Vincent, 2002). The light activation of channelrhodopsin in the thalamic neurons likely overcomes these intrinsic cell properties, eliminating the slow after-hyperpolarization and the corresponding decrease in firing rate and synchronization of the thalamic population. However, further experiments are needed to more fully understand this process in vivo and to better understand the interaction between intrinsic properties of thalamic neurons and the ascending input from the brainstem.

The adaptation in firing rate across VPm and S1 evolved on similar timescales as reflected qualitatively in Figure 1E and quantitatively through relatively comparable time constants, although the time constant associated with the slower component was longer in S1 as compared with VPm. However, the extent of the adaptation on this timescale was more extreme in S1 as compared with VPm, as reflected in the steady-state firing rate relative to the nonadapted firing rate at the onset of the adaptation, consistent with previous observations (Chung et al., 2002; Khatri et al., 2004; Gabernet et al., 2005). Further, within S1 the RS neurons adapted more than the FS neurons, also consistent with previous reports (Wright et al., 2021). It is also the case that despite having a similar qualitative nature, the slower timescale component of the thalamic synchrony is faster than that of the thalamic firing rate. This could suggest that the adaptation of the VPm synchrony arises from a different mechanism from that of the VPm firing rate, but it could also be the case that any relationship between firing rate and synchrony would be nonlinear and not expected to evolve in the same way. In any case, given the extreme sensitivity of the cortex to the synchrony of the thalamic input (Bruno and Sakmann, 2006; Wang et al., 2010b; Wright et al., 2021), even the small decrease in thalamic synchrony observed on the longer timescale here does not preclude thalamic synchrony from being a primary role player in the corresponding cortical firing rate adaptation, as suggested by the optogenetic control experiments.

In this study, we were able to capture multiple timescales of adaptation using a 10 Hz repetitive stimulation over 10 s. It is important to note, however, that adaptation is a change in function that takes time to develop and also dissipate. It has been shown that the temporal scales over which adaptive dynamics buildup and recover are distinct and also highly dependent upon the duration of exposure to repetitive stimulation (Dragoi et al., 2000; Chung et al., 2002; Patterson et al., 2013). In general, brief exposure produces adaptation that develops and recovers rapidly (Bonds, 1991; Nelson, 1991; Müller et al., 1999), while more prolonged stimulation can result in slower and more lasting forms of changes (Greenlee et al., 1991). Additionally, thalamic and cortical neurons also exhibit diverse timescales of responsiveness to repeated stimulation (Chung et al., 2002). Therefore, as a rule of thumb, an intertrial interval of at least four times the duration of adapting stimulus train (10 s) was utilized in this study, while acquiring at least 100 trials of repeated measurements of the adaptation dynamics. As a control, we also assessed the difference between the magnitude of the response to the first deflection of the adapting train at the beginning and end of a stimulus block. We did not observe a significant difference in this measure, suggesting that the majority of the neurons returned to the baseline condition prior to the initiation of the next trial, although longitudinal effects cannot be completely ruled out. For a subset of the data, we also probed the recovery for each thalamic and cortical neuron and found that most of the cells showed a full recovery to the baseline condition. Finally, it is important to note that different aspects of the adapting stimulus itself, such as frequency content and velocity, can strongly shape the nature of the adaptation as we have previously shown (Zheng et al., 2015). It is likely that the adaptation observed here varies in detail with the adapting stimulus properties. Further studies are needed to fully explore the range of sensory stimulus conditions that give rise to the phenomena described here and how broadly the results here generalize.

The direct assessment of the transformation of adaptation across the thalamocortical junction in this study is facilitated by our simultaneous multichannel electrode recordings of putatively connected pairs of thalamocortical neurons. From the total number of topographically aligned thalamocortical pairs recorded in this study, we identified ∼33% as putatively connected, similar to but slightly lower than that previously described in the cat visual pathway (42%; Reid and Alonso, 1995). This difference in inferred connectivity can likely be attributed to the conservative nature of our inference methodology, which was explicitly designed to tolerate misses over false alarms (Liew et al., 2021). Further, the inference of connectivity is very sensitive to the number of spikes used in the analysis; the nature of the multielectrode arrays used in this study resulted in sampling of neurons with highly diverse firing rates, including those with relatively low firing rates that precluded the inference of connectivity. These two factors likely resulted in excluding pairs that were actually connected from the analyses here. When parsed based on S1 cell-type classification, we further found a higher probability of likely connectivity among VPm→FS (46.8%) as compared with VPm→RS (24.8%) pairs, consistent with previous findings (Bruno and Simons, 2002; Bruno and Sakmann, 2006).

Given that this study was performed in the somatosensory pathway of isoflurane-anesthetized mice, it is important to acknowledge some of the limitations in the findings. The large majority of in vivo studies on the relative role of short-term depression of thalamocortical and intracortical synapses have been conducted under anesthesia (Heiss et al., 2008; Chung et al., 2019), and the current study cannot rule out the potential influence of TC synaptic depression in the early, fast timescale component of adaptation. In contrast, the adaptive changes in synchronous thalamic firing have recently been reported to play a more prominent role than synaptic depression in the TC and IC synapse under wakefulness (Wright et al., 2021). One possibility is that the overlap between distinct mechanisms during wakefulness is smaller compared with the condition under anesthesia. It was reported that the thalamocortical circuit may be in an elevated baseline firing state during wakefulness (Castro-Alamancos, 2004a), hence the suggestion that TC synapses might already be depressed in this state. When TC synaptic depression is fully engaged during the high thalamic firing rates associated with wakefulness, thalamic desynchronization may dominate and exert an even greater effect on regulating the nature of information conveyed downstream to cortical targets. Further, the observed timescales of the adaptive processes are likely different under anesthesia. However, to fully disentangle these effects and the relative contribution from each mechanism, further techniques would need to be employed, such as selective manipulation of TC and CC synaptic depression combined with simultaneous, multisite recordings, and ultimately would need to be conducted in the absence of anesthesia.

Finally, the results here suggest that although multiple mechanisms may interact and transition dynamically at various timescales, adaptation over the timecourse that we examined in this study improves theoretical discrimination at the expense of detection. This similarity in function supports the possibility of a unifying framework across disparate mechanisms and timescales, which is to allow the brain to maintain a flexible representation in response to stimulus statistics that span a wide range in our natural environment. Moreover, these results further support Barlow's efficient coding hypothesis such that the reduction in responsiveness of neuronal response in the TC circuit is not a negative attribute that signifies a loss of sensitivity to stimuli, but instead a mechanism that could mediate an adaptive trade-off in coding strategy to put the brain, in this case the primary sensory cortex, in a state that optimizes neural coding efficiency (Barlow, 1961).

Previous Section Next Section

Footnotes

Received June 6, 2024.
Revision received January 3, 2025.
Accepted January 19, 2025.

This work was supported by NIH National Institute of Neurological Disorders and Stroke (NINDS) BRAIN Grant R01NS104928 and NINDS BRAIN Grant RF1NS128896. Y.J.L. was supported by a Georgia Tech-Emory-PKU Global Biomedical Engineering Fellowship. E.D.D. was supported by a National Science Foundation Graduate Research Fellowship and the Howard Hughes Medical Institute through the James H. Gilliam Fellowships for Advanced Study program. Y.Z. was supported by STI2030-Major Projects 2022ZD0204900. We thank Eunji Cheong and Cheong lab members at Yonsei University for helpful conversations at various stages of this work.
↵*Y.J.L. and E.D.D. contributed equally to this work.
The authors declare no competing financial interests.
Correspondence should be addressed to Garrett B. Stanley at garrett.stanley{at}bme.gatech.edu.

Copyright © 2025 the authors

SfN exclusive license.

Previous Section

References

↵
1. Adibi M,
2. Clifford CWG,
3. Arabzadeh E
(2013) Informational basis of sensory adaptation: entropy and single-spike efficiency in rat barrel cortex. J Neurosci 33:14921–14926. https://doi.org/10.1523/JNEUROSCI.1313-13.2013 pmid:24027291
Abstract/FREE Full Text
↵
1. Adibi M,
2. Lampl I
(2021) Sensory adaptation in the whisker-mediated tactile system: physiology, theory, and function. Front Neurosci 15:770011. https://doi.org/10.3389/fnins.2021.770011 pmid:34776857
CrossRef Medline
↵
1. Ahissar E,
2. Sosnik R,
3. Haidarliu S
(2000) Transformation from temporal to rate coding in a somatosensory thalamocortical pathway. Nature 406:302–306. https://doi.org/10.1038/35018568
CrossRef Medline
↵
1. Rosenblith WA
1. Barlow H
(1961) Possible principles underlying the transformations of sensory messages. In: Sensory communication (Rosenblith WA, ed), pp 217–234. Cambridge, MA: MIT Presshttps://doi.org/10.7551/mitpress/9780262518420.003.0013.
Google Scholar
↵
1. Barthó P,
2. Hirase H,
3. Monconduit L,
4. Zugaro M,
5. Harris KD,
6. Buzsáki G
(2004) Characterization of neocortical principal cells and interneurons by network interactions and extracellular features. J Neurophysiol 92:600–608. https://doi.org/10.1152/jn.01170.2003
CrossRef Medline
↵
1. Beierlein M,
2. Connors BW
(2002) Short-term dynamics of thalamocortical and intracortical synapses onto layer 6 neurons in neocortex. J Neurophysiol 88:1924–1932. https://doi.org/10.1152/jn.2002.88.4.1924
CrossRef Medline
↵
1. Bolus MF,
2. Willats AA,
3. Rozell CJ,
4. Stanley GB
(2021) State-space optimal feedback control of optogenetically driven neural activity. J Neural Eng 18:10.1088/1741-2552/abb89c. https://doi.org/10.1088/1741-2552/abb89c pmid:32932241
Medline Google Scholar
↵
1. Bonds AB
(1991) Temporal dynamics of contrast gain in single cells of the cat striate cortex. Vis Neurosci 6:239–255. https://doi.org/10.1017/S0952523800006258
CrossRef Medline
↵
1. Borden PY,
2. Ortiz AD,
3. Waiblinger C,
4. Sederberg AJ,
5. Morrissette AE,
6. Forest CR,
7. Jaeger D,
8. Stanley GB
(2017) Genetically expressed voltage sensor ArcLight for imaging large scale cortical activity in the anesthetized and awake mouse. Neurophotonics 4:031212. https://doi.org/10.1117/1.NPh.4.3.031212 pmid:28491905
CrossRef Medline
↵
1. Borden PY,
2. Wright NC,
3. Morrissette AE,
4. Jaeger D,
5. Haider B,
6. Stanley GB
(2022) Thalamic bursting and the role of timing and synchrony in thalamocortical signaling in the awake mouse. Neuron 110:2836–2853.e8. https://doi.org/10.1016/j.neuron.2022.06.008 pmid:35803270
CrossRef Medline
↵
1. Bruno RM,
2. Sakmann B
(2006) Cortex is driven by weak but synchronously active thalamocortical synapses. Science 312:1622–1627. https://doi.org/10.1126/science.1124593
Abstract/FREE Full Text
↵
1. Bruno RM,
2. Simons DJ
(2002) Feedforward mechanisms of excitatory and inhibitory cortical receptive fields. J Neurosci 22:10966–10975. https://doi.org/10.1523/JNEUROSCI.22-24-10966.2002 pmid:12486192
Abstract/FREE Full Text
↵
1. Castro-Alamancos MA
(2004a) Absence of rapid sensory adaptation in neocortex during information processing states. Neuron 41:455–464. https://doi.org/10.1016/S0896-6273(03)00853-5
CrossRef Medline
↵
1. Castro-Alamancos MA
(2004b) Dynamics of sensory thalamocortical synaptic networks during information processing states. Prog Neurobiol 74:213–247. https://doi.org/10.1016/j.pneurobio.2004.09.002
CrossRef Medline
↵
1. Chung JE, et al.
(2019) High-density, long-lasting, and multi-region electrophysiological recordings using polymer electrode arrays. Neuron 101:21–31.e5. https://doi.org/10.1016/j.neuron.2018.11.002 pmid:30502044
CrossRef Medline
↵
1. Chung S,
2. Li X,
3. Nelson SB
(2002) Short-term depression at thalamocortical synapses contributes to rapid adaptation of cortical sensory responses in vivo. Neuron 34:437–446. https://doi.org/10.1016/S0896-6273(02)00659-1
CrossRef Medline
↵
1. Cruikshank SJ,
2. Lewis TJ,
3. Connors BW
(2007) Synaptic basis for intense thalamocortical activation of feedforward inhibitory cells in neocortex. Nat Neurosci 10:462–468. https://doi.org/10.1038/nn1861
CrossRef Medline
↵
1. Dimwamwa ED,
2. Pala A,
3. Chundru V,
4. Wright NC,
5. Stanley GB
(2023) Dynamic corticothalamic modulation of the somatosensory thalamocortical circuit during wakefulness. Biorxiv.
Google Scholar
↵
1. Dragoi V,
2. Sharma J,
3. Sur M
(2000) Adaptation-induced plasticity of orientation tuning in adult visual cortex. Neuron 28:287–298. https://doi.org/10.1016/S0896-6273(00)00103-3
CrossRef Medline
↵
1. Drew PJ,
2. Abbott LF
(2006) Models and properties of power-law adaptation in neural systems. J Neurophysiol 96:826–833. https://doi.org/10.1152/jn.00134.2006
CrossRef Medline
↵
1. Fairhall A,
2. Lewen G,
3. Bialek W,
4. van Steveninck R
(2000) Multiple timescales of adaptation in a neural code. In: advances in neural information processing systems. Denver: MIT Press.
Google Scholar
↵
1. Franklin K,
2. Paxinos G
(2008) The mouse brain in stereotaxic coordinates, compact. The coronal plates and diagrams. Amsterdam: Elsevier Academic Press.
Google Scholar
↵
1. Gabernet L,
2. Jadhav SP,
3. Feldman DE,
4. Carandini M,
5. Scanziani M
(2005) Somatosensory integration controlled by dynamic thalamocortical feed-forward inhibition. Neuron 48:315–327. https://doi.org/10.1016/j.neuron.2005.09.022
CrossRef Medline
↵
1. Ganmor E,
2. Katz Y,
3. Lampl I
(2010) Intensity-dependent adaptation of cortical and thalamic neurons is controlled by brainstem circuits of the sensory pathway. Neuron 66:273–286. https://doi.org/10.1016/j.neuron.2010.03.032
CrossRef Medline
↵
1. Gerfen CR,
2. Paletzki R,
3. Heintz N
(2013) GENSAT BAC cre-recombinase driver lines to study the functional organization of cerebral cortical and basal ganglia circuits. Neuron 80:1368–1383. https://doi.org/10.1016/j.neuron.2013.10.016 pmid:24360541
CrossRef Medline Google Scholar
↵
1. Goaillard J-M,
2. Vincent P
(2002) Serotonin suppresses the slow afterhyperpolarization in rat intralaminar and midline thalamic neurones by activating 5-HT7 receptors. J Physiol 541:453–465. https://doi.org/10.1113/jphysiol.2001.013896 pmid:12042351
CrossRef Medline
↵
1. Greenlee MW,
2. Georgeson MA,
3. Magnussen S,
4. Harris JP
(1991) The time course of adaptation to spatial contrast. Vision Res 31:223–236. https://doi.org/10.1016/0042-6989(91)90113-J
CrossRef Medline
↵
1. Heiss JE,
2. Katz Y,
3. Ganmor E,
4. Lampl I
(2008) Shift in the balance between excitation and inhibition during sensory adaptation of S1 neurons. J Neurosci 28:13320–13330. https://doi.org/10.1523/JNEUROSCI.2646-08.2008 pmid:19052224
Abstract/FREE Full Text
↵
1. Katz Y,
2. Heiss JE,
3. Lampl I
(2006) Cross-whisker adaptation of neurons in the rat barrel cortex. J Neurosci 26:13363–13372. https://doi.org/10.1523/JNEUROSCI.4056-06.2006 pmid:17182787
Abstract/FREE Full Text
↵
1. Khatri V,
2. Hartings JA,
3. Simons DJ
(2004) Adaptation in thalamic barreloid and cortical barrel neurons to periodic whisker deflections varying in frequency and velocity. J Neurophysiol 92:3244–3254. https://doi.org/10.1152/jn.00257.2004
CrossRef Medline
↵
1. Kheradpezhouh E,
2. Adibi M,
3. Arabzadeh E
(2017) Response dynamics of rat barrel cortex neurons to repeated sensory stimulation. Sci Rep 7:11445. https://doi.org/10.1038/s41598-017-11477-6 pmid:28904406
CrossRef Medline
↵
1. Latimer KW,
2. Barbera D,
3. Sokoletsky M,
4. Awwad B,
5. Katz Y,
6. Nelken I,
7. Lampl I,
8. Fairhall AL,
9. Priebe NJ
(2019) Multiple timescales account for adaptive responses across sensory cortices. J Neurosci 39:10019–10033. https://doi.org/10.1523/JNEUROSCI.1642-19.2019 pmid:31662427
Abstract/FREE Full Text
↵
1. Latimer KW,
2. Fairhall AL
(2020) Capturing multiple timescales of adaptation to second-order statistics with generalized linear models: gain scaling and fractional differentiation. Front Syst Neurosci 14:60. https://doi.org/10.3389/fnsys.2020.00060 pmid:33013331
Medline
↵
1. Lefort S,
2. Tomm C,
3. Floyd Sarria J-C,
4. Petersen CCH
(2009) The excitatory neuronal network of the C2 barrel column in mouse primary somatosensory cortex. Neuron 61:301–316. https://doi.org/10.1016/j.neuron.2008.12.020
CrossRef Medline
↵
1. Liew YJ,
2. Pala A,
3. Whitmire CJ,
4. Stoy WA,
5. Forest CR,
6. Stanley GB
(2021) Inferring thalamocortical monosynaptic connectivity in vivo. J Neurophysiol 125:2408–2431. https://doi.org/10.1152/jn.00591.2020 pmid:33978507
CrossRef Medline
↵
1. Ljung L
(1987) System identification: theory for the user. Upper Saddle River, NJ: Prentice Hall.
Google Scholar
↵
1. Lundstrom BN,
2. Fairhall AL,
3. Maravall M
(2010) Multiple timescale encoding of slowly varying whisker stimulus envelope in cortical and thalamic neurons in vivo. J Neurosci 30:5071–5077. https://doi.org/10.1523/JNEUROSCI.2193-09.2010 pmid:20371827
Abstract/FREE Full Text
↵
1. Mainen ZF,
2. Sejnowski TJ
(1995) Reliability of spike timing in neocortical neurons. Science 268:1503–1506. https://doi.org/10.1126/science.7770778
Abstract/FREE Full Text
↵
1. Maravall M,
2. Alenda A,
3. Bale MR,
4. Petersen RS
(2013) Transformation of adaptation and gain rescaling along the whisker sensory pathway. PLoS One 8:e82418. https://doi.org/10.1371/journal.pone.0082418 pmid:24349279
CrossRef Medline
↵
1. Maravall M,
2. Petersen RS,
3. Fairhall AL,
4. Arabzadeh E,
5. Diamond ME
(2007) Shifts in coding properties and maintenance of information transmission during adaptation in barrel cortex. PLoS Biol 5:e19. https://doi.org/10.1371/journal.pbio.0050019 pmid:17253902
CrossRef Medline
↵
1. Masri R,
2. Bezdudnaya T,
3. Trageser JC,
4. Keller A
(2008) Encoding of stimulus frequency and sensor motion in the posterior medial thalamic nucleus. J Neurophysiol 100:681–689. https://doi.org/10.1152/jn.01322.2007 pmid:18234976
CrossRef Medline
↵
1. Mohar B,
2. Ganmor E,
3. Lampl I
(2015) Faithful representation of tactile intensity under different contexts emerges from the distinct adaptive properties of the first somatosensory relay stations. J Neurosci 35:6997–7002. https://doi.org/10.1523/JNEUROSCI.4358-14.2015 pmid:25948252
Abstract/FREE Full Text
↵
1. Mohar B,
2. Katz Y,
3. Lampl I
(2013) Opposite adaptive processing of stimulus intensity in two major nuclei of the somatosensory brainstem. J Neurosci 33:15394–15400. https://doi.org/10.1523/JNEUROSCI.1886-13.2013 pmid:24068807
Abstract/FREE Full Text
↵
1. Müller JR,
2. Metha AB,
3. Krauskopf J,
4. Lennie P
(1999) Rapid adaptation in visual cortex to the structure of images. Science 285:1405–1408. https://doi.org/10.1126/science.285.5432.1405
Abstract/FREE Full Text
↵
1. Nelson SB
(1991) Temporal interactions in the cat visual system. III. Pharmacological studies of cortical suppression suggest a presynaptic mechanism. J Neurosci 11:369–380. https://doi.org/10.1523/JNEUROSCI.11-02-00369.1991 pmid:1992007
Abstract/FREE Full Text
↵
1. Ollerenshaw DR,
2. Zheng HJV,
3. Millard DC,
4. Wang Q,
5. Stanley GB
(2014) The adaptive trade-off between detection and discrimination in cortical representations and behavior. Neuron 81:1152–1164. https://doi.org/10.1016/j.neuron.2014.01.025 pmid:24607233
CrossRef Medline
↵
1. Pachitariu M,
2. Sridhar S,
3. Pennington J,
4. Stringer C
(2024) Spike sorting with Kilosort4. Nat Methods 21:914–921. https://doi.org/10.1038/s41592-024-02232-7.
CrossRef Medline Google Scholar
↵
1. Pala A,
2. Petersen CCH
(2015) In vivo measurement of cell-type-specific synaptic connectivity and synaptic transmission in layer 2/3 mouse barrel cortex. Neuron 85:68–75. https://doi.org/10.1016/j.neuron.2014.11.025 pmid:25543458
CrossRef Medline
↵
1. Patterson CA,
2. Wissig SC,
3. Kohn A
(2013) Distinct effects of brief and prolonged adaptation on orientation tuning in primary visual cortex. J Neurosci 33:532–543. https://doi.org/10.1523/JNEUROSCI.3345-12.2013 pmid:23303933
Abstract/FREE Full Text
↵
1. Pinto DJ,
2. Brumberg JC,
3. Simons DJ
(2000) Circuit dynamics and coding strategies in rodent somatosensory cortex. J Neurophysiol 83:1158–1166. https://doi.org/10.1152/jn.2000.83.3.1158
CrossRef Medline
↵
1. Reid RC,
2. Alonso JM
(1995) Specificity of monosynaptic connections from thalamus to visual cortex. Nature 378:281–284. https://doi.org/10.1038/378281a0
CrossRef Medline
↵
1. Reig R,
2. Gallego R,
3. Nowak LG,
4. Sanchez-Vives MV
(2006) Impact of cortical network activity on short-term synaptic depression. Cereb Cortex 16:688–695. https://doi.org/10.1093/cercor/bhj014
CrossRef Medline
↵
1. Rossant C,
2. Harris KD
(2013) Semi-automatic spike sorting with high-count channel probes. BMC Neurosci 14(Suppl 1):P160. https://doi.org/10.1186/1471-2202-14-S1-P160
Google Scholar
↵
1. Simons DJ,
2. Carvell GE
(1989) Thalamocortical response transformation in the rat vibrissa/barrel system. J Neurophysiol 61:311–330. https://doi.org/10.1152/jn.1989.61.2.311
CrossRef Medline
↵
1. Storchi R,
2. Bale MR,
3. Biella GEM,
4. Petersen RS
(2012) Comparison of latency and rate coding for the direction of whisker deflection in the subcortical somatosensory pathway. J Neurophysiol 108:1810–1821. https://doi.org/10.1152/jn.00921.2011 pmid:22815402
CrossRef Medline
↵
1. Swadlow HA,
2. Gusev AG
(2001) The impact of “bursting” thalamic impulses at a neocortical synapse. Nat Neurosci 4:402–408. https://doi.org/10.1038/86054
CrossRef Medline
↵
1. Wang H-P,
2. Spencer D,
3. Fellous J-M,
4. Sejnowski TJ
(2010a) Synchrony of thalamocortical inputs maximizes cortical reliability. Science 328:106–109. https://doi.org/10.1126/science.1183108 pmid:20360111
Abstract/FREE Full Text
↵
1. Wang Q,
2. Webber RM,
3. Stanley GB
(2010b) Thalamic synchrony and the adaptive gating of information flow to cortex. Nat Neurosci 13:1534–1541. https://doi.org/10.1038/nn.2670 pmid:21102447
CrossRef Medline
↵
1. Whitmire CJ,
2. Stanley GB
(2016) Rapid sensory adaptation redux: a circuit perspective. Neuron 92:298–315. https://doi.org/10.1016/j.neuron.2016.09.046 pmid:27764664
CrossRef Medline
↵
1. Wright NC,
2. Borden PY,
3. Liew YJ,
4. Bolus MF,
5. Stoy WM,
6. Forest CR,
7. Stanley GB
(2021) Rapid cortical adaptation and the role of thalamic synchrony during wakefulness. J Neurosci 41:5421–5439. https://doi.org/10.1523/JNEUROSCI.3018-20.2021 pmid:33986072
Abstract/FREE Full Text
↵
1. Zheng HJV,
2. Wang Q,
3. Stanley GB
(2015) Adaptive shaping of cortical response selectivity in the vibrissa pathway. J Neurophysiol 113:3850–3865. https://doi.org/10.1152/jn.00978.2014 pmid:25787959
CrossRef Medline

[1] ↵

Adibi M,

Clifford CWG,

Arabzadeh E

(2013) Informational basis of sensory adaptation: entropy and single-spike efficiency in rat barrel cortex. J Neurosci 33:14921–14926. https://doi.org/10.1523/JNEUROSCI.1313-13.2013 pmid:24027291

Abstract/FREE Full Text

[2] Adibi M,

[3] Clifford CWG,

[4] Arabzadeh E

[5] ↵

Adibi M,

Lampl I

(2021) Sensory adaptation in the whisker-mediated tactile system: physiology, theory, and function. Front Neurosci 15:770011. https://doi.org/10.3389/fnins.2021.770011 pmid:34776857

CrossRef Medline

[6] Adibi M,

[7] Lampl I

[8] ↵

Ahissar E,

Sosnik R,

Haidarliu S

(2000) Transformation from temporal to rate coding in a somatosensory thalamocortical pathway. Nature 406:302–306. https://doi.org/10.1038/35018568

CrossRef Medline

[9] Ahissar E,

[10] Sosnik R,

[11] Haidarliu S

[12] ↵

Rosenblith WA

Barlow H

(1961) Possible principles underlying the transformations of sensory messages. In: Sensory communication (Rosenblith WA, ed), pp 217–234. Cambridge, MA: MIT Presshttps://doi.org/10.7551/mitpress/9780262518420.003.0013.

Google Scholar

[13] Rosenblith WA

[14] Barlow H

[15] ↵

Barthó P,

Hirase H,

Monconduit L,

Zugaro M,

Harris KD,

Buzsáki G

(2004) Characterization of neocortical principal cells and interneurons by network interactions and extracellular features. J Neurophysiol 92:600–608. https://doi.org/10.1152/jn.01170.2003

CrossRef Medline

[16] Barthó P,

[17] Hirase H,

[18] Monconduit L,

[19] Zugaro M,

[20] Harris KD,

[21] Buzsáki G

[22] ↵

Beierlein M,

Connors BW

(2002) Short-term dynamics of thalamocortical and intracortical synapses onto layer 6 neurons in neocortex. J Neurophysiol 88:1924–1932. https://doi.org/10.1152/jn.2002.88.4.1924

CrossRef Medline

[23] Beierlein M,

[24] Connors BW

[25] ↵

Bolus MF,

Willats AA,

Rozell CJ,

Stanley GB

(2021) State-space optimal feedback control of optogenetically driven neural activity. J Neural Eng 18:10.1088/1741-2552/abb89c. https://doi.org/10.1088/1741-2552/abb89c pmid:32932241

Medline Google Scholar

[26] Bolus MF,

[27] Willats AA,

[28] Rozell CJ,

[29] Stanley GB

[30] ↵

Bonds AB

(1991) Temporal dynamics of contrast gain in single cells of the cat striate cortex. Vis Neurosci 6:239–255. https://doi.org/10.1017/S0952523800006258

CrossRef Medline

[31] Bonds AB

[32] ↵

Borden PY,

Ortiz AD,

Waiblinger C,

Sederberg AJ,

Morrissette AE,

Forest CR,

Jaeger D,

Stanley GB

(2017) Genetically expressed voltage sensor ArcLight for imaging large scale cortical activity in the anesthetized and awake mouse. Neurophotonics 4:031212. https://doi.org/10.1117/1.NPh.4.3.031212 pmid:28491905

CrossRef Medline

[33] Borden PY,

[34] Ortiz AD,

[35] Waiblinger C,

[36] Sederberg AJ,

[37] Morrissette AE,

[38] Forest CR,

[39] Jaeger D,

[40] Stanley GB

[41] ↵

Borden PY,

Wright NC,

Morrissette AE,

Jaeger D,

Haider B,

Stanley GB

(2022) Thalamic bursting and the role of timing and synchrony in thalamocortical signaling in the awake mouse. Neuron 110:2836–2853.e8. https://doi.org/10.1016/j.neuron.2022.06.008 pmid:35803270

CrossRef Medline

[42] Borden PY,

[43] Wright NC,

[44] Morrissette AE,

[45] Jaeger D,

[46] Haider B,

[47] Stanley GB

[48] ↵

Bruno RM,

Sakmann B

(2006) Cortex is driven by weak but synchronously active thalamocortical synapses. Science 312:1622–1627. https://doi.org/10.1126/science.1124593

Abstract/FREE Full Text

[49] Bruno RM,

[50] Sakmann B

[51] ↵

Bruno RM,

Simons DJ

(2002) Feedforward mechanisms of excitatory and inhibitory cortical receptive fields. J Neurosci 22:10966–10975. https://doi.org/10.1523/JNEUROSCI.22-24-10966.2002 pmid:12486192

Abstract/FREE Full Text

[52] Bruno RM,

[53] Simons DJ

[54] ↵

Castro-Alamancos MA

(2004a) Absence of rapid sensory adaptation in neocortex during information processing states. Neuron 41:455–464. https://doi.org/10.1016/S0896-6273(03)00853-5

CrossRef Medline

[55] Castro-Alamancos MA

[56] ↵

Castro-Alamancos MA

(2004b) Dynamics of sensory thalamocortical synaptic networks during information processing states. Prog Neurobiol 74:213–247. https://doi.org/10.1016/j.pneurobio.2004.09.002

CrossRef Medline

[57] Castro-Alamancos MA

[58] ↵

Chung JE, et al.

(2019) High-density, long-lasting, and multi-region electrophysiological recordings using polymer electrode arrays. Neuron 101:21–31.e5. https://doi.org/10.1016/j.neuron.2018.11.002 pmid:30502044

CrossRef Medline

[59] Chung JE, et al.

[60] ↵

Chung S,

Li X,

Nelson SB

(2002) Short-term depression at thalamocortical synapses contributes to rapid adaptation of cortical sensory responses in vivo. Neuron 34:437–446. https://doi.org/10.1016/S0896-6273(02)00659-1

CrossRef Medline

[61] Chung S,

[62] Li X,

[63] Nelson SB

[64] ↵

Cruikshank SJ,

Lewis TJ,

Connors BW

(2007) Synaptic basis for intense thalamocortical activation of feedforward inhibitory cells in neocortex. Nat Neurosci 10:462–468. https://doi.org/10.1038/nn1861

CrossRef Medline

[65] Cruikshank SJ,

[66] Lewis TJ,

[67] Connors BW

[68] ↵

Dimwamwa ED,

Pala A,

Chundru V,

Wright NC,

Stanley GB

(2023) Dynamic corticothalamic modulation of the somatosensory thalamocortical circuit during wakefulness. Biorxiv.

Google Scholar

[69] Dimwamwa ED,

[70] Pala A,

[71] Chundru V,

[72] Wright NC,

[73] Stanley GB

[74] ↵

Dragoi V,

Sharma J,

Sur M

(2000) Adaptation-induced plasticity of orientation tuning in adult visual cortex. Neuron 28:287–298. https://doi.org/10.1016/S0896-6273(00)00103-3

CrossRef Medline

[75] Dragoi V,

[76] Sharma J,

[77] Sur M

[78] ↵

Drew PJ,

Abbott LF

(2006) Models and properties of power-law adaptation in neural systems. J Neurophysiol 96:826–833. https://doi.org/10.1152/jn.00134.2006

CrossRef Medline

[79] Drew PJ,

[80] Abbott LF

[81] ↵

Fairhall A,

Lewen G,

Bialek W,

van Steveninck R

(2000) Multiple timescales of adaptation in a neural code. In: advances in neural information processing systems. Denver: MIT Press.

Google Scholar

[82] Fairhall A,

[83] Lewen G,

[84] Bialek W,

[85] van Steveninck R

[86] ↵

Franklin K,

Paxinos G

(2008) The mouse brain in stereotaxic coordinates, compact. The coronal plates and diagrams. Amsterdam: Elsevier Academic Press.

Google Scholar

[87] Franklin K,

[88] Paxinos G

[89] ↵

Gabernet L,

Jadhav SP,

Feldman DE,

Carandini M,

Scanziani M

(2005) Somatosensory integration controlled by dynamic thalamocortical feed-forward inhibition. Neuron 48:315–327. https://doi.org/10.1016/j.neuron.2005.09.022

CrossRef Medline

[90] Gabernet L,

[91] Jadhav SP,

[92] Feldman DE,

[93] Carandini M,

[94] Scanziani M

[95] ↵

Ganmor E,

Katz Y,

Lampl I

(2010) Intensity-dependent adaptation of cortical and thalamic neurons is controlled by brainstem circuits of the sensory pathway. Neuron 66:273–286. https://doi.org/10.1016/j.neuron.2010.03.032

CrossRef Medline

[96] Ganmor E,

[97] Katz Y,

[98] Lampl I

[99] ↵

Gerfen CR,

Paletzki R,

Heintz N

(2013) GENSAT BAC cre-recombinase driver lines to study the functional organization of cerebral cortical and basal ganglia circuits. Neuron 80:1368–1383. https://doi.org/10.1016/j.neuron.2013.10.016 pmid:24360541

CrossRef Medline Google Scholar

[100] Gerfen CR,

[101] Paletzki R,

[102] Heintz N

[103] ↵

Goaillard J-M,

Vincent P

(2002) Serotonin suppresses the slow afterhyperpolarization in rat intralaminar and midline thalamic neurones by activating 5-HT7 receptors. J Physiol 541:453–465. https://doi.org/10.1113/jphysiol.2001.013896 pmid:12042351

CrossRef Medline

[104] Goaillard J-M,

[105] Vincent P

[106] ↵

Greenlee MW,

Georgeson MA,

Magnussen S,

Harris JP

(1991) The time course of adaptation to spatial contrast. Vision Res 31:223–236. https://doi.org/10.1016/0042-6989(91)90113-J

CrossRef Medline

[107] Greenlee MW,

[108] Georgeson MA,

[109] Magnussen S,

[110] Harris JP

[111] ↵

Heiss JE,

Katz Y,

Ganmor E,

Lampl I

(2008) Shift in the balance between excitation and inhibition during sensory adaptation of S1 neurons. J Neurosci 28:13320–13330. https://doi.org/10.1523/JNEUROSCI.2646-08.2008 pmid:19052224

Abstract/FREE Full Text

[112] Heiss JE,

[113] Katz Y,

[114] Ganmor E,

[115] Lampl I

[116] ↵

Katz Y,

Heiss JE,

Lampl I

(2006) Cross-whisker adaptation of neurons in the rat barrel cortex. J Neurosci 26:13363–13372. https://doi.org/10.1523/JNEUROSCI.4056-06.2006 pmid:17182787

Abstract/FREE Full Text

[117] Katz Y,

[118] Heiss JE,

[119] Lampl I

[120] ↵

Khatri V,

Hartings JA,

Simons DJ

(2004) Adaptation in thalamic barreloid and cortical barrel neurons to periodic whisker deflections varying in frequency and velocity. J Neurophysiol 92:3244–3254. https://doi.org/10.1152/jn.00257.2004

CrossRef Medline

[121] Khatri V,

[122] Hartings JA,

[123] Simons DJ

[124] ↵

Kheradpezhouh E,

Adibi M,

Arabzadeh E

(2017) Response dynamics of rat barrel cortex neurons to repeated sensory stimulation. Sci Rep 7:11445. https://doi.org/10.1038/s41598-017-11477-6 pmid:28904406

CrossRef Medline

[125] Kheradpezhouh E,

[126] Adibi M,

[127] Arabzadeh E

[128] ↵

Latimer KW,

Barbera D,

Sokoletsky M,

Awwad B,

Katz Y,

Nelken I,

Lampl I,

Fairhall AL,

Priebe NJ

(2019) Multiple timescales account for adaptive responses across sensory cortices. J Neurosci 39:10019–10033. https://doi.org/10.1523/JNEUROSCI.1642-19.2019 pmid:31662427

Abstract/FREE Full Text

[129] Latimer KW,

[130] Barbera D,

[131] Sokoletsky M,

[132] Awwad B,

[133] Katz Y,

[134] Nelken I,

[135] Lampl I,

[136] Fairhall AL,

[137] Priebe NJ

[138] ↵

Latimer KW,

Fairhall AL

(2020) Capturing multiple timescales of adaptation to second-order statistics with generalized linear models: gain scaling and fractional differentiation. Front Syst Neurosci 14:60. https://doi.org/10.3389/fnsys.2020.00060 pmid:33013331

Medline

[139] Latimer KW,

[140] Fairhall AL

[141] ↵

Lefort S,

Tomm C,

Floyd Sarria J-C,

Petersen CCH

(2009) The excitatory neuronal network of the C2 barrel column in mouse primary somatosensory cortex. Neuron 61:301–316. https://doi.org/10.1016/j.neuron.2008.12.020

CrossRef Medline

[142] Lefort S,

[143] Tomm C,

[144] Floyd Sarria J-C,

[145] Petersen CCH

[146] ↵

Liew YJ,

Pala A,

Whitmire CJ,

Stoy WA,

Forest CR,

Stanley GB

(2021) Inferring thalamocortical monosynaptic connectivity in vivo. J Neurophysiol 125:2408–2431. https://doi.org/10.1152/jn.00591.2020 pmid:33978507

CrossRef Medline

[147] Liew YJ,

[148] Pala A,

[149] Whitmire CJ,

[150] Stoy WA,

[151] Forest CR,

[152] Stanley GB

[153] ↵

Ljung L

(1987) System identification: theory for the user. Upper Saddle River, NJ: Prentice Hall.

Google Scholar

[154] Ljung L

[155] ↵

Lundstrom BN,

Fairhall AL,

Maravall M

(2010) Multiple timescale encoding of slowly varying whisker stimulus envelope in cortical and thalamic neurons in vivo. J Neurosci 30:5071–5077. https://doi.org/10.1523/JNEUROSCI.2193-09.2010 pmid:20371827

Abstract/FREE Full Text

[156] Lundstrom BN,

[157] Fairhall AL,

[158] Maravall M

[159] ↵

Mainen ZF,

Sejnowski TJ

(1995) Reliability of spike timing in neocortical neurons. Science 268:1503–1506. https://doi.org/10.1126/science.7770778

Abstract/FREE Full Text

[160] Mainen ZF,

[161] Sejnowski TJ

[162] ↵

Maravall M,

Alenda A,

Bale MR,

Petersen RS

(2013) Transformation of adaptation and gain rescaling along the whisker sensory pathway. PLoS One 8:e82418. https://doi.org/10.1371/journal.pone.0082418 pmid:24349279

CrossRef Medline

[163] Maravall M,

[164] Alenda A,

[165] Bale MR,

[166] Petersen RS

[167] ↵

Maravall M,

Petersen RS,

Fairhall AL,

Arabzadeh E,

Diamond ME

(2007) Shifts in coding properties and maintenance of information transmission during adaptation in barrel cortex. PLoS Biol 5:e19. https://doi.org/10.1371/journal.pbio.0050019 pmid:17253902

CrossRef Medline

[168] Maravall M,

[169] Petersen RS,

[170] Fairhall AL,

[171] Arabzadeh E,

[172] Diamond ME

[173] ↵

Masri R,

Bezdudnaya T,

Trageser JC,

Keller A

(2008) Encoding of stimulus frequency and sensor motion in the posterior medial thalamic nucleus. J Neurophysiol 100:681–689. https://doi.org/10.1152/jn.01322.2007 pmid:18234976

CrossRef Medline

[174] Masri R,

[175] Bezdudnaya T,

[176] Trageser JC,

[177] Keller A

[178] ↵

Mohar B,

Ganmor E,

Lampl I

(2015) Faithful representation of tactile intensity under different contexts emerges from the distinct adaptive properties of the first somatosensory relay stations. J Neurosci 35:6997–7002. https://doi.org/10.1523/JNEUROSCI.4358-14.2015 pmid:25948252

Abstract/FREE Full Text

[179] Mohar B,

[180] Ganmor E,

[181] Lampl I

[182] ↵

Mohar B,

Katz Y,

Lampl I

(2013) Opposite adaptive processing of stimulus intensity in two major nuclei of the somatosensory brainstem. J Neurosci 33:15394–15400. https://doi.org/10.1523/JNEUROSCI.1886-13.2013 pmid:24068807

Abstract/FREE Full Text

[183] Mohar B,

[184] Katz Y,

[185] Lampl I

[186] ↵

Müller JR,

Metha AB,

Krauskopf J,

Lennie P

(1999) Rapid adaptation in visual cortex to the structure of images. Science 285:1405–1408. https://doi.org/10.1126/science.285.5432.1405

Abstract/FREE Full Text

[187] Müller JR,

[188] Metha AB,

[189] Krauskopf J,

[190] Lennie P

[191] ↵

Nelson SB

(1991) Temporal interactions in the cat visual system. III. Pharmacological studies of cortical suppression suggest a presynaptic mechanism. J Neurosci 11:369–380. https://doi.org/10.1523/JNEUROSCI.11-02-00369.1991 pmid:1992007

Abstract/FREE Full Text

[192] Nelson SB

[193] ↵

Ollerenshaw DR,

Zheng HJV,

Millard DC,

Wang Q,

Stanley GB

(2014) The adaptive trade-off between detection and discrimination in cortical representations and behavior. Neuron 81:1152–1164. https://doi.org/10.1016/j.neuron.2014.01.025 pmid:24607233

CrossRef Medline

[194] Ollerenshaw DR,

[195] Zheng HJV,

[196] Millard DC,

[197] Wang Q,

[198] Stanley GB

[199] ↵

Pachitariu M,

Sridhar S,

Pennington J,

Stringer C

(2024) Spike sorting with Kilosort4. Nat Methods 21:914–921. https://doi.org/10.1038/s41592-024-02232-7.

CrossRef Medline Google Scholar

[200] Pachitariu M,

[201] Sridhar S,

[202] Pennington J,

[203] Stringer C

[204] ↵

Pala A,

Petersen CCH

(2015) In vivo measurement of cell-type-specific synaptic connectivity and synaptic transmission in layer 2/3 mouse barrel cortex. Neuron 85:68–75. https://doi.org/10.1016/j.neuron.2014.11.025 pmid:25543458

CrossRef Medline

[205] Pala A,

[206] Petersen CCH

[207] ↵

Patterson CA,

Wissig SC,

Kohn A

(2013) Distinct effects of brief and prolonged adaptation on orientation tuning in primary visual cortex. J Neurosci 33:532–543. https://doi.org/10.1523/JNEUROSCI.3345-12.2013 pmid:23303933

Abstract/FREE Full Text

[208] Patterson CA,

[209] Wissig SC,

[210] Kohn A

[211] ↵

Pinto DJ,

Brumberg JC,

Simons DJ

(2000) Circuit dynamics and coding strategies in rodent somatosensory cortex. J Neurophysiol 83:1158–1166. https://doi.org/10.1152/jn.2000.83.3.1158

CrossRef Medline

[212] Pinto DJ,

[213] Brumberg JC,

[214] Simons DJ

[215] ↵

Reid RC,

Alonso JM

(1995) Specificity of monosynaptic connections from thalamus to visual cortex. Nature 378:281–284. https://doi.org/10.1038/378281a0

CrossRef Medline

[216] Reid RC,

[217] Alonso JM

[218] ↵

Reig R,

Gallego R,

Nowak LG,

Sanchez-Vives MV

(2006) Impact of cortical network activity on short-term synaptic depression. Cereb Cortex 16:688–695. https://doi.org/10.1093/cercor/bhj014

CrossRef Medline

[219] Reig R,

[220] Gallego R,

[221] Nowak LG,

[222] Sanchez-Vives MV

[223] ↵

Rossant C,

Harris KD

(2013) Semi-automatic spike sorting with high-count channel probes. BMC Neurosci 14(Suppl 1):P160. https://doi.org/10.1186/1471-2202-14-S1-P160

Google Scholar

[224] Rossant C,

[225] Harris KD

[226] ↵

Simons DJ,

Carvell GE

(1989) Thalamocortical response transformation in the rat vibrissa/barrel system. J Neurophysiol 61:311–330. https://doi.org/10.1152/jn.1989.61.2.311

CrossRef Medline

[227] Simons DJ,

[228] Carvell GE

[229] ↵

Storchi R,

Bale MR,

Biella GEM,

Petersen RS

(2012) Comparison of latency and rate coding for the direction of whisker deflection in the subcortical somatosensory pathway. J Neurophysiol 108:1810–1821. https://doi.org/10.1152/jn.00921.2011 pmid:22815402

CrossRef Medline

[230] Storchi R,

[231] Bale MR,

[232] Biella GEM,

[233] Petersen RS

[234] ↵

Swadlow HA,

Gusev AG

(2001) The impact of “bursting” thalamic impulses at a neocortical synapse. Nat Neurosci 4:402–408. https://doi.org/10.1038/86054

CrossRef Medline

[235] Swadlow HA,

[236] Gusev AG

[237] ↵

Wang H-P,

Spencer D,

Fellous J-M,

Sejnowski TJ

(2010a) Synchrony of thalamocortical inputs maximizes cortical reliability. Science 328:106–109. https://doi.org/10.1126/science.1183108 pmid:20360111

Abstract/FREE Full Text

[238] Wang H-P,

[239] Spencer D,

[240] Fellous J-M,

[241] Sejnowski TJ

[242] ↵

Wang Q,

Webber RM,

Stanley GB

(2010b) Thalamic synchrony and the adaptive gating of information flow to cortex. Nat Neurosci 13:1534–1541. https://doi.org/10.1038/nn.2670 pmid:21102447

CrossRef Medline

[243] Wang Q,

[244] Webber RM,

[245] Stanley GB

[246] ↵

Whitmire CJ,

Stanley GB

(2016) Rapid sensory adaptation redux: a circuit perspective. Neuron 92:298–315. https://doi.org/10.1016/j.neuron.2016.09.046 pmid:27764664

CrossRef Medline

[247] Whitmire CJ,

[248] Stanley GB

[249] ↵

Wright NC,

Borden PY,

Liew YJ,

Bolus MF,

Stoy WM,

Forest CR,

Stanley GB

(2021) Rapid cortical adaptation and the role of thalamic synchrony during wakefulness. J Neurosci 41:5421–5439. https://doi.org/10.1523/JNEUROSCI.3018-20.2021 pmid:33986072

Abstract/FREE Full Text

[250] Wright NC,

[251] Borden PY,

[252] Liew YJ,

[253] Bolus MF,

[254] Stoy WM,

[255] Forest CR,

[256] Stanley GB

[257] ↵

Zheng HJV,

Wang Q,

Stanley GB

(2015) Adaptive shaping of cortical response selectivity in the vibrissa pathway. J Neurophysiol 113:3850–3865. https://doi.org/10.1152/jn.00978.2014 pmid:25787959

CrossRef Medline

[258] Zheng HJV,

[259] Wang Q,

[260] Stanley GB

The Journal of NeuroscienceSociety for Neuroscience

Multiple Distinct Timescales of Rapid Sensory Adaption in the Thalamocortical Circuit

Abstract

Significance Statement

Introduction

Materials and Methods

Animals

Head-plate implantation

Intrinsic signal optical imaging of the mouse S1

Anesthetized electrophysiology

VPm recordings

S1 recordings

Barreloid/barrel targeting

Sensory stimulus

Adapting stimulus

Weak sinusoidal stimulus

Optogenetic stimulation of neural activity

Postmortem histology

Analytical methods

Spike sorting

Analysis of sensory response

VPm neuron validation

Cortical neuron classification

Principal whisker (PW) identification

Determining topographic alignment

Characterization of the adaptation dynamics with exponential models

Analysis of velocity sensitivity

Spike cross-correlation analysis

Inferring monosynaptic connectivity

Measuring synaptic efficacy

Measuring VPm synchrony

Receiver operator characteristic (ROC) analysis

Statistical analyses

Experimental design

Data accessibility

Results

Differential rapid sensory adaptation in VPm and S1

Multiple timescales underlie the adaptation dynamics of cortical neurons in response to repetitive sensory stimulation

Multiple timescales underlie the adaptation dynamics of thalamic firing rate and population synchrony in response to repetitive sensory stimulation

Multiple timescales underlie the adaptation dynamics of thalamocortical synaptic efficacy in response to repetitive stimulation

Direct optogenetic drive of VPm thalamus eliminates longer timescale adaptation in VPm

Optogenetic drive of VPm results in a loss of longer timescale adaptation in S1

Adaptation over multiple timescales degrades stimulus detection for an ideal observer of cortical activity

Adaptation over multiple timescales enhances stimulus discrimination for an ideal observer of cortical activity

Discussion

Footnotes

References

Navigate This Article

Google Scholar

PubMed

ORCID

Social Bookmarking

The Journal of Neuroscience Society for Neuroscience