Abstract

The structure of local circuits is highly conserved across the cortex, yet the spatial and temporal properties of population activity differ fundamentally in sensory-level and association-level areas. In the sensory cortex, population activity has a shorter timescale and decays sharply over distance, supporting a population code for the fine-scale features of sensory stimuli. In the association cortex, population activity has a longer timescale and spreads over wider distances, a code that is suited to holding information in memory and driving behavior. We tested whether these differences in activity dynamics could be explained by differences in network structure. We targeted photostimulations to single excitatory neurons of layer 2/3, while monitoring surrounding population activity using two-photon calcium imaging. Experiments were performed in the auditory (AC) and posterior parietal cortex (PPC) within the same mice of both sexes, which also expressed a red fluorophore in somatostatin-expressing interneurons (SOM). In both cortical regions, photostimulations resulted in a spatially restricted zone of positive influence on neurons closely neighboring the targeted neuron and a more spatially diffuse zone of negative influence affecting more distant neurons (akin to a network-level “suppressive surround”). However, the relative spatial extents of positive and negative influence were different in AC and PPC. In PPC, the central zone of positive influence was wider, but the negative suppressive surround was more narrow than in AC, which could account for the larger-scale network dynamics in PPC. The more narrow central positive influence zone and wider suppressive surround in AC could serve to sharpen sensory representations.

Significance Statement

The structure of local circuits is conserved across the cortex, and yet local processing goals vary across the sensorimotor hierarchy, from sensory perception to the control of behavior. It has been unclear whether these differences in function require different network organization. To probe the spatial structure of networks in sensory and association-level cortex, we photostimulated single excitatory neurons and measured the effects on local population activity in mice. Stimulations triggered a centered, positive activity change in neighboring neurons and a surrounding, suppressive change in more distant neurons. The relative sizes of the center and surround differed across areas, suggesting that network structure is tailored for sharper, more restricted activity in the sensory cortex and more dense network activity in the association cortex.

Introduction

Cortical circuits consist of a rich diversity of cell types, which interact to transform incoming signals by shaping receptive field properties (Adesnik et al., 2012; Lee et al., 2012; Wilson et al., 2012), controlling response gain (Atallah et al., 2012; Wilson et al., 2012; Veit et al., 2022), and amplifying (Ko et al., 2011; Cossell et al., 2015; Akitake et al., 2023; Oldenburg et al., 2024) or sparsening responses to stimuli (Chettih and Harvey, 2019; Oldenburg et al., 2024). Circuit dissection experiments in the sensory cortex have begun to characterize these local network interactions and their corresponding computational impacts (Yoshimura and Callaway, 2005; Yoshimura et al., 2005; Fino and Yuste, 2011; Packer and Yuste, 2011; Chettih and Harvey, 2019; Akitake et al., 2023; Oldenburg et al., 2024). Excitatory neurons form functionally specific recurrent subnetworks in the sensory cortex, where neurons with similar response properties tend to form stronger, bidirectional synaptic connections (Yoshimura et al., 2005; Ko et al., 2011; Perin et al., 2011; Li et al., 2012; Cossell et al., 2015). In vivo, two-photon optogenetic stimulation of single neurons combined with simultaneous calcium imaging of the local population can be used to characterize “influence maps” of single neurons’ activity on the surrounding population. These influence maps are the net result of triggered activation across excitatory and inhibitory connections and their synaptic weights across the network, possibly even including longer-range interactions with unobserved neurons in different layers or brain regions. Although this approach generally cannot reveal the actual synaptic pathways from the stimulated neuron to the neurons it affects, the influence map represents the functional, network-level impact of a single neuron on the activity in the local population (Kwan and Dan, 2012; Chettih and Harvey, 2019; Oldenburg et al., 2024). This in vivo influence mapping approach, applied in primary sensory cortex, suggests that recurrent excitation can serve as signal amplification or to sparsen stimulus-evoked activity, depending on the details of experimental conditions, such as the number of neurons being stimulated, their tuning preferences, and the sensory context (Chettih and Harvey, 2019; Akitake et al., 2023; Oldenburg et al., 2024). The functional impact of recurrent excitatory interactions is thus state dependent in sensory cortex.

Inhibitory interneurons are key in modulating both local network states and the impacts of feedforward and recurrent excitation. Through cell-type–specific interactions (Yoshimura and Callaway, 2005; Lee et al., 2012; Wilson et al., 2012; Pfeffer et al., 2013; Pi et al., 2013; Green et al., 2023), inhibitory interneurons can profoundly shift population activity dynamics and also shape the response properties of individual neurons. For example, somatostatin-expressing (SOM) interneurons directly inhibit both other inhibitory interneurons and excitatory pyramidal cells (Pfeffer et al., 2013; Condylis et al., 2022), refining receptive fields (Adesnik et al., 2012; Wilson et al., 2012; Kato et al., 2017), decorrelating population activity at slow timescales (Chen et al., 2015), and modulating high-frequency oscillations (Veit et al., 2017). SOM neurons tend to be highly coordinated as populations (Karnani et al., 2016; Knoblich et al., 2019; Khoury et al., 2022), and so they are particularly well suited to modulate local network states in the cortex.

Previously, cellular and synaptic-level circuit manipulations have characterized the organization of local circuits and the local influence of individual neurons in the cortex almost exclusively in sensory regions. So it is not yet known how the structural and functional organization of local circuits may differ across cortical regions with different computational goals. In sensory cortex, topographically organized inputs from core sensory nuclei of the thalamus generate or maintain tuning for specific stimulus features, such as retinotopic location or the orientation of an edge in the primary visual cortex (Reid and Alonso, 1995) or best frequency across the tonotopic axis in the primary auditory cortex. These tuning preferences are then further shaped by local excitatory and inhibitory interactions (Somers et al., 1998; Troyer et al., 1998). In association-level areas such as the posterior parietal cortex (PPC), sensory inputs across several modalities are flexibly combined (Olcese et al., 2013; Song et al., 2017; Mohan et al., 2018) toward sensorimotor decisions (Shadlen and Newsome, 1996; Harvey et al., 2012; Tseng et al., 2022; Zhou et al., 2023). The computations that integrate sensory information across modalities and across time in areas like PPC may depend on more spatially distributed networks compared with those needed to encode fine-scale sensory features in topographically organized sensory areas (Markowitz et al., 2015; Safavi et al., 2018). Systematic changes in neuronal morphology (Elston, 2000) and in the timescale of activity in individual neurons (Murray et al., 2014) and in populations (Honey et al., 2012; Runyan et al., 2017; Safavi et al., 2018; Khoury et al., 2022) across the cortical hierarchy have led to hypotheses that the contribution of recurrent connections to neuronal activity becomes systematically stronger across the cortical hierarchy, in support of cognitive processes requiring temporal integration (Chaudhuri et al., 2015; Hasson et al., 2015).

We hypothesized that the density and spatial scale of recurrent connections could explain differing spatial and temporal scales of activity reported across the cortex and across cell types (Khoury et al., 2022). If this is the case, the activation of single excitatory neurons should influence the local population more widely in the association cortex than in the sensory cortex. To test this, we targeted photostimulations to individual excitatory neurons in the auditory cortex (AC) and posterior parietal cortex (PPC) while monitoring spike-related activity in neighboring excitatory (E) and SOM neurons. In both regions, the stimulations positively influenced a narrow, central zone of neighboring neurons, and negatively influenced a diffuse, surrounding zone including more distant neurons. The relative sizes of these two zones differed across areas, reflecting the observed population activity differences in AC and PPC. In PPC, the central positive zone of influence was wider, and the surrounding negative zone was more narrow than in AC. As a consequence, excitation can spread more easily across the network in PPC, while in AC, the spread of excitation is restricted by the stronger suppressive surround.

Materials and MethodsExperimental model and subject details

All procedures were approved by the University of Pittsburgh Institutional Animal Care and Use Committee. Homozygous SOM-Cre mice (Sst-IRES-Cre, JAX Stock #013044) were crossed with homozygous Ai14 mice (RCL-tdT-D, JAX Stock #007914) obtained from the Jackson Laboratory, and all experiments were performed in the F1 generation, which expressed tdTomato in SOM+ neurons. Mice were group housed in cages with between two and four mice. Adult (8–24 weeks) male and female mice were used for experiments (4 male, 6 female). Mice were housed on a reversed 12 h light/dark cycle, and all experiments were performed in the dark (active) phase.

Surgery

Mice were anesthetized with isoflurane (4% for induction, and 1–2% maintenance during surgery) and mounted on a stereotaxic frame (David Kopf Instruments). Ophthalmic ointment was applied to cover the eyes (Henry Schein). Dexamethasone was injected 12–24 h prior to surgery, and carprofen and dexamethasone (Covetrus) were injected subcutaneously immediately prior to surgery for pain management and to reduce the inflammatory response. Two 2 × 2 mm craniotomies were made, over left AC (centered on the temporal ridge, with the posterior edge along the lambdoid suture) and PPC (centered at 2 mm posterior and 1.75 mm lateral to the bregma). One to four evenly spaced ∼60 nl injections of a viral mixture were made in each cranial window, centered on the same coordinates listed above. The viral mixture contained AAV1-synapsin-l-GCamp6f (Addgene, MA stock #100837; Chen et al., 2013) and AAV9-CAMKII-mScarlet-C1V1-KV2.1 (Addgene, MA stock # 124650), mixed 1:1 and diluted to a titer of ∼1 × 1012 vg/ml using sterile PBS, and a micromanipulator (QUAD, Sutter) was used to target injections ∼250 µm under the dura at each site, where ∼60 nl virus was pressure-injected over 5–10 min. pAAV-CamKIIa-C1V1(t/t)-mScarlet-KV2.1 was a gift from Christopher Harvey (Addgene viral prep # 124650-AAV9; http://n2t.net/addgene:124650; RRID:Addgene_124650; Chettih and Harvey, 2019). Pipettes were not removed until 5 min postinjection to prevent backflow. Dental cement (Parkell) sealed a glass coverslip (3 mm) over a drop of Kwik-Sil (World Precision Instruments) on the craniotomy. Using dental cement, a one-sided titanium headplate was attached to the right hemisphere of the skull. After mice had recovered from the anesthesia, they were returned to their home cages and received oral carprofen tablets (Fisher Scientific) daily for 3 d postsurgery.

Two-photon microscope

Imaging and photostimulation were performed entirely within layer 2/3, using a resonant scanning two-photon microscope (Ultima 2Pplus, Bruker). Images were collected at a 30 Hz frame rate and 512 × 512 pixel resolution through a 16× water immersion lens (Nikon CF175, 16X/0.8 NA). On separate days in the same mice, either AC or PPC was imaged at a depth between 150 and 300 µm, corresponding to layers 2/3 of the cortex. For AC imaging, the objective was rotated 35–45° from vertical, and for PPC imaging, it was rotated to 5–15° from vertical, matching the angle of the cranial window implant. AC and PPC imaging was performed on alternating days in the same mice.

Excitation light was provided by a femtosecond infrared laser (InSight X3, Spectra-Physics) tuned to 920 nm. Photostimulation was achieved by controlling a 1,045 nm beam (secondary output from the InSight X3) with an independent set of galvanometers. The 920 and 1,045 nm beams were combined with a dichroic (Chroma, ZT1040crb). Green and red wavelength emission light was separated through a 565 nm low-pass filter before passing through bandpass filters (Chroma, ET525/70 and ET595/50). PrairieView software (v5.5, Bruker) was used to control the microscope.

Photostimulation

Excitatory neurons expressing mScarlet could be distinguished from the tdTomato + SOM neurons by the subcellular localization of the fluorophores as well as by collecting images at 800 nm excitation, at which mScarlet is more strongly excited than tdTomato (Drobizhev et al., 2009). A total of 16–47 mScarlet+ neurons were selected as targets in each experiment. The number depended on expression levels, as cells were chosen to have robust mScarlet and GCaMP expression levels. Then, 5–10 “control” targets not expressing mScarlet were also chosen, so that the specificity of stimulation effects could be assessed. The secondary pair of galvanometers was used to direct spiral scans of the 1,045 nm beam over individual neurons.

Spirals (13 µm in diameter) were directed to each target at 250 Hz for 100 ms, at 20–50 mW. These parameters were chosen to elicit reliable responses in the targeted neurons, while limiting stimulation power (Chettih and Harvey, 2019; Russell et al., 2022; Oldenburg et al., 2024). At the beginning of each experiment, one target was chosen for stimulation at a range of excitation powers. The lowest power that could reliably activate the cell was selected for the rest of the experimental session. For influence mapping, the full set of individual mScarlet+ and control targets were stimulated in pseudorandom order with a 1 s interstimulus interval. The 1 s interstimulus interval was chosen to allow for a large number of trial repeats within the imaging session, while also allowing for enough time for the neural responses elicited by each stimulation to return to baseline before the next stimulation. At least 100 trial repeats were performed, as this is the number of stimulations that could be accomplished within a typical 60–90 min imaging session. Different pseudorandom orderings were used across trial repeats, to avoid driving plasticity-related changes in network structure.

Behavioral monitoring

Throughout the imaging experiments, mice ran voluntarily on a spherical treadmill. Running velocity was monitored on pitch, roll, and yaw axes using two optical sensors (ADNS-98000, Tindie) held adjacent to the spherical treadmill. A microcontroller (Teensy, 3.1, Adafruit) communicated with the sensors, demixing their inputs to produce one output channel per rotational axis using custom code. Outputs controlling the galvanometers were synchronized with running velocity using a digital oscilloscope (WaveSurfer).

Sound stimuli

In auditory cortex imaging sessions, either immediately before or following influence mapping, the same field of view was imaged while presenting sinusoid amplitude-modulated (10 Hz) pure tone stimuli. One magnetic speaker was pointed to the ear contralateral to the imaging hemisphere (MF1-S, Tucker-Davis). Tones were played at frequencies of 4, 8, 12, 16, 24, and 32 kHz for 1 s each. Sounds were played at a single intensity (70 dB). Sound responsiveness of each neuron was calculated based on the mean z-scored deconvolved activity of each neuron aligned on sound onset. We chose not to measure sound frequency tuning in the posterior parietal cortex (PPC) because of its unreliable sound-evoked activity in passive listening contexts.

Image processing

For each field of view, the raw calcium movies were concatenated for motion correction, cell body identification, and fluorescence and neuropil extraction. These processing steps were performed using Suite2p 0.9.3 in Python (Pachitariu et al., 2017). Suite2p first registered images to account for brain motion, and clustered neighboring pixels with similar time courses into regions of interest (ROIs). ROIs were manually curated using the Suite2p GUI, to ensure that only cell bodies as opposed to dendritic processes were included in analysis, based on morphology. Cells expressing tdTomato (SOM cells) were identified using a threshold applied in the Suite2p GUI based on mean fluorescence in the red channel after bleed-through correction applied by Suite2p's cell detection algorithm, along with manual correction. For each ROI, Suite2p returned a raw fluorescence timeseries, as well as an estimate of neuropil fluorescence that could contaminate the signal. For each cell, we scaled the neuropil fluorescence by a factor by 0.7 and subtracted this timeseries from the ROI's raw fluorescence timeseries to obtain a neuropil-corrected fluorescence signal for each selected cell.

ΔF/F and deconvolution

Once the neuropil-corrected fluorescence was obtained for each neuron, we calculated ΔF/F for each cell in each imaging frame by calculating (F − Fbaseline)/Fbaseline for each frame, where F is the fluorescence of a given cell at that frame and Fbaseline is the eighth percentile of that cell's fluorescence spanning 450 frames before and after (∼15 s each way, 30 s total). ΔF/F timeseries were then deconvolved to estimate the relative spike rate in each imaging frame using the OASIS toolbox (Friedrich et al., 2017). We used the AR1 FOOPSI algorithm and allowed the toolbox to optimize the convolution kernel, baseline fluorescence, and noise distribution. A threshold of 0.05 a.u. was applied to remove all events with low magnitude from deconvolved activity timeseries. All analyses were performed with both ΔF/F and deconvolved activity and showed the same trends. Results in the figures are based on ΔF/F.

Pairwise noise correlations

Before computing pairwise noise correlations, each neuron's trial-averaged response to a given target was subtracted from single trial responses in the 333 ms after the target stimulation onset. These values were then correlated between each target neuron and other neuron in the population, but only during trials that the particular target neuron was not stimulated. The goal was to obtain the pairwise correlations of the neurons outside of times that either neuron was stimulated, to measure their natural cofluctuations in activity.

Influence calculation

Influence of each target on each other neuron was calculated using deconvolved spike rates as well as dF/F, with similar results. The mean prestimulus response of each neuron (binned across 10 imaging frames, or 333 ms, prior to the target's stimulation) was subtracted from its mean poststimulus response (in the 10 frames, or 333 ms, immediately following the target's stimulation onset). These time windows were chosen to limit the noise in the baseline measurement and to capture the majority of each neuron's response to the stimulation. This value was then normalized by the standard deviation of this difference across all trials. To relate influence level to running speed, the prestimulus running speed was binned into quartiles for each imaging session separately. Influence was then compared across running speed quartiles.

Histology

After all imaging sessions had been acquired, each mouse was transcardially perfused with saline and then 4% paraformaldehyde. The brain was extracted, cryoprotected, embedded, frozen, and sliced. Once slide mounted, we stained brains with DAPI to be able to identify anatomical landmarks. We used anatomical structure to verify the locations of the viral injections in AC and PPC.

Experimental design and statistical analysis

Unless otherwise stated, pairwise comparisons were done with two-sided paired or unpaired permutation (i.e., randomization) tests with 10,000 iterations, where p = 0.0001 indicates the highest significance achievable given the number of runs performed. All permutation tests were performed for differences in means. Permutation tests were chosen to avoid making unnecessary assumptions about the normality of the data. When multiple comparisons were made between groups, p values were Bonferroni-corrected to assess significance. To compare the effects of brain region, cell type, and intersomatic distance on influence, we performed a three-way ANOVA. Post hoc tests were calculated as multiple comparisons using the Tukey–Kramer test. Full values and statistical details from all figures are thoroughly summarized in Tables 1 and 2 and Extended Data Table 2-1.

Table 1.

Related to Figure 2

Table 2.

Related to Figure 8

ResultsMapping the influence of single excitatory neurons in vivo

Our goal was to compare the functional structure of recurrent networks in layer 2/3 of the auditory and parietal cortex, where population activity dynamics differ dramatically across areas (Runyan et al., 2017; Valente et al., 2021; Khoury et al., 2022). We expressed a red shifted opsin (C1V1) in excitatory (E) neurons, a red fluorophore (tdTomato) in SOM neurons, and a green calcium indicator (GCaMP6f) in all neurons in both the auditory cortex (AC) and posterior parietal cortex (PPC) of 10 SOM-Cre × ai14 mice (Fig. 1A,B). During imaging sessions (25 AC sessions and 30 PPC sessions, AC and PPC were imaged within the same mice on alternating days), mice ran voluntarily on a spherical treadmill. To map the influence of E neurons on local SOM and Non-SOM neurons in AC and PPC, we performed spiral scans with a 1,045 nm laser beam over individual C1V1-expressing E neurons to trigger action potentials (Rickgauer and Tank, 2009; Chettih and Harvey, 2019) while simultaneously raster scanning a 920 nm laser beam to monitor spike-related calcium activity in GCaMP-expressing neurons. In a typical imaging session, we targeted spirals (100 ms in duration) to ∼30 individual E neurons for at least 100 trial repeats, in pseudorandom order, with 1 s between stimulations. Successfully stimulated neurons were, by definition, excitatory neurons, as they expressed C1V1 under the CaMKII promoter. The other neurons in the field of view were sorted based on their tdTomato expression: tdTomato+ neurons were considered SOM, and tdTomato− neurons were considered Non-SOM, which can include Non-SOM inhibitory interneurons, but the majority would be excitatory neurons. Each field of view contained 14.8 ± 6.6 SOM neurons (mean ± standard deviation, here and onward) and 222.9 ± 96.0 Non-SOM neurons in AC and PPC. Of the 30.6 ± 7.4 neurons targeted for photostimulation, 22.6 ± 6.9 were successfully stimulated. All imaging and stimulations were targeted to similar depths from the cortical surface (120–300 µm).

Figure 1.

Mapping the influence of single E neurons in vivo. A, SOM-Cre × ai14 mice ran voluntarily on a spherical treadmill throughout experiments. A speaker was used to present pure tones to characterize auditory responses. Excitation for imaging was provided by a laser tuned to 920 nm. Excitation for stimulation was provided by a 1,045 nm laser that was controlled by an independent scan path. B, An excitatory opsin (C1V1) was virally expressed in excitatory neurons. tdTomato was expressed transgenically in SOM neurons, and GCaMP was expressed in all neurons virally, in SOM-Cre x ai14 mice. Left, Example field of view imaged at 920 nm, visualizing GCaMP (green) and tdTomato (red). Right, Same field of view, imaged at 780 nm, where the mScarlet (red, also expressed with C1V1) but not tdTomato can be visualized. C, Left, Example influence map obtained from the field of view in B. Excited neuron's location indicated by the red “spiral.” Significantly positively influenced neurons are circled in black, negatively influenced neurons in magenta. Neurons that were not significantly influenced by the target's stimulation are circled in gray. Right, Example trial-averaged responses of significantly positively influenced E cells, negatively influenced E cells, and positively influenced SOM cells when this neuron was stimulated. D, As in C, for another neuron that was targeted in the same session and field of view. E, Control experiment to determine the spatial resolution of the photostimulations. A neuron was stimulated and then spirals were targeted to adjacent locations, as indicated by the red spirals superimposed on the image. Right, Trial-averaged responses of the neuron during spiral scans that were directed to each location indicated on the left. The neuron was only significantly responsive when the spiral was focused directly on it. We only included other neurons that were at least 25 µm away from the spiral in our influence measurements. According to our tests, this is a conservative enough approach because we rarely observed direct stimulation of a neuron even 12.5 µm away. Shading: SEM.

The “influence” of each successfully stimulated neuron on every other simultaneously imaged neuron was then assessed. We defined “influence” of a photostimulation target on another neuron as the other neuron's mean response to the target's stimulation (response during 1 s after the target's stimulation onset minus response in the ∼300 ms before stimulation onset) across all trials, divided by its standard deviation. We successfully stimulated 506 E neurons in AC, testing 125,496 potential influences on Non-SOM and 6,846 potential influences on SOM neurons. In PPC, we successfully stimulated 577 E neurons, testing 171,692 potential influences on Non-SOM neurons and 12,152 potential influences on SOM neurons in PPC. On average, a sparse subset of neighboring neurons were positively influenced (7.9% ± 8.7% of all neurons, defined as influence greater than 99% of trial shuffled data, see Materials and Methods, considering only neurons at least 25 µm away from the target), and a smaller subset of local neurons were negatively influenced (3.4% ± 3.5% of all neurons, defined as influence less than 99% of trial shuffled data; Fig. 1C,D).

To estimate the spatial specificity of the spiral stimulation, in most sessions we selected a single C1V1-expressing neuron that was on the edge of the viral injection site. We then systematically stimulated the location of the neuron's soma, and neighboring distances at 12.5 µm increments (Fig. 1E). For most targets, significant responses were only evoked when the spirals were targeted within 12.5 µm of the soma. To be conservative, in subsequent analyses, neurons within 25 µm of the stimulation target were excluded.

Center/surround organization of local influence of single E neurons

To display and analyze the spatial spread of E neurons' local network influence in each area, we centered each targeted neuron and overlaid and summed the relative spatial positions of significantly positively or negatively influenced neurons, normalizing by the total number of targeted neurons (Fig. 2A–D). Visually, the spatial extent of positive influence was restricted and centered around the stimulated neuron in both cortical regions, though this zone of positive influence appeared wider in PPC than in AC (Fig. 2A,B). In contrast, the spatial pattern of negative influence was more wide and diffuse than positive influence and appeared more dominant in AC than PPC (Fig. 2C,D).

Figure 2.

The spatial spread of excitatory influence in AC and PPC. A, Maps of influence on Non-SOM neurons, across all targeted E neurons and all fields of view in AC (left) and PPC (right). Each targeted neuron was centered, and the pixel locations of its significantly positively influenced neurons set to 1. Then, all 506 AC maps and 577 PPC maps were summed and divided by the total number of targeted neurons, so that the color at each pixel location indicates the number of significantly positively influenced Non-SOM neurons per targeted neuron, at each relative location. B, As in A but for positive influence on SOM neurons (n's of targeted neurons are the same as in A). C, D, As in A and B but for negative influence. E, Cumulative distributions of the intersomatic distances between the targets and significantly influenced Non-SOM cells, in AC (red) and PPC (blue). Gray, distributions of distances of all Non-SOM neurons to the target. Solid, positively influenced neurons; dashed, negatively influenced neurons. See Table 1 for n's. F, As in E, but showing the distributions of distances of SOM neurons to the targets. G, Same data as in E and F, but plotting the mean influence on all neurons, binned by distance to the target. In all panels, only neurons at least 25 µm from the target were considered. See Table 1 and Extended Data Table 2-1 for full values and statistics.

To quantify the differences in the spatial spread of positive and negative influence of E cells in AC and PPC, we measured the intersomatic distance from each targeted neuron to every other neuron in the field of view. We then compared the distributions of intersomatic distances of significantly positively and negatively influenced Non-SOM (Fig. 2E) and SOM (Fig. 2F) neurons. In both regions, positively influenced neurons tended to be closer to the targeted neurons than both negatively influenced and the full distribution of imaged neurons (AC Non-SOM: positively influenced vs negatively influenced p = 0.000999; PPC Non-SOM: p = 0.000999; AC SOM: p = 0.000999; PPC SOM: p = 0.0001; see Table 1 for full values and statistics).

To test our hypothesis that recurrent excitation spreads more widely in PPC than in AC, we then compared the distribution of distances to positively influenced neurons across areas. Consistent with our hypothesis, the zone of positively influenced neurons was more spatially restricted in AC than in PPC: the distances from positively influenced cells to the targeted neurons were longer in PPC than AC (Fig. 2A,E,G; p = 0.00099). Similarly, distances to positively influenced SOM neurons were longer in PPC than in AC (Fig. 2B,F,G; p = 0.00070). However, the distance to negatively influenced Non-SOM neurons was longer in AC than PPC (Fig. 2C; p = 0.000999). The distance to negatively influenced SOM neurons did not differ in AC and PPC (p = 0.4203).

To simultaneously test for effects of brain region, cell type, and intersomatic distance on the targets' influence, we binned the intersomatic distances from targets to other neurons and performed a three-way ANOVA (Fig. 2G; results reported in Extended Data Table 2-1). There were significant main effects of brain area and intersomatic distance (p < 0.0001), and significant interactions between brain area and intersomatic distance (p < 0.0001) and between cell type and distance (p = 0.0072). Post hoc tests revealed greater influence on both SOM and Non-SOM neurons in PPC compared with AC neurons at distances <100 µm (p = 9.59 × 10−6; Tukey–Kramer test; for multiple comparisons, see Extended Data Table 2-2 for full list of statistical comparisons).

To summarize, stimulation of E cells resulted in a central zone of positive influence on the local population, with a diffuse “surround” zone of negative influence. Positive influence spread farther in PPC, but negative influence spread farther in AC.

The effects of single E neuron stimulation on local population activity

Our previous analysis focused on the effects of target stimulation on other individual neurons. To capture the population-level impact of the targeted neuron's stimulation, we next studied the population pattern of local influence in “population activity space.” We reasoned that if the local recurrent excitatory synapses are denser and connect neurons over wider distances in PPC, this should also be evident in the pattern of neurons in each local population that is positively versus negatively affected by the target neuron's activity. In high dimensional space, where each axis is defined as the activity of one neuron in the population (n neurons in a population define an n-dimensional space), we identified a “stimulus axis” for each stimulated neuron (Fig. 3A). This is the axis in n-dimensional space that connects the population's average prestimulus activity to the population's response to the stimulation of that neuron. When we projected population activity onto this axis, using both SOM and Non-SOM neurons, a robust stimulus-evoked population response was revealed (Fig. 3B). Next, we examined the distribution of weights of the nonstimulated neurons that defined each “stimulus axis.” The mean stimulus axis weight was more positive in PPC than AC (AC: 0.018  ±  0.080, PPC: 0.026  ±  0.089, p = 0.0001, permutation test), suggesting that while the density of E influence in the two regions is sparse, influence on the local population was more positive in PPC than in AC (Fig. 3C–E).

Figure 3.

Excitatory neuron influence on the local population. A, A “stimulus axis” was defined in high dimensional neural activity space for the population's response to each specific target's stimulation that best aligned with the population's response to the target stimulation. B, The population mean response in the stimulus axis (defined independently for each target stimulus) is plotted for AC and PPC. n = 506 AC populations and 577 PPC populations. C, The distribution of weights along the stimulus axis, of all AC neurons at least 25 µm from the targets (n = 132,259 neurons). D, As in B, for PPC neurons (n = 183,730 neurons). E, Same data as in B and C, but plotted as bar plots of the mean stimulus axis weight across all AC and all PPC neurons. Error bars indicate standard error of the mean. PPC weights were significantly more positive than AC weights (p = 1.00 × 10−4, permutation test). F, G, The positively (F) and negatively (G) weighted neurons in each stimulus axis were separated and binned by distance to the stimulated neuron in AC and PPC. Asterisks indicate Bonferroni-corrected significance in unpaired permutation tests comparing the weights between AC and PPC in each distance bin. ***p < 0.001, **p < 0.01, *p < 0.05.

To consider the relationship between stimulus axis weight and distance to the stimulated target neuron, we next split all nontargeted neurons into positively (Fig. 3F) and negatively weighted subsets (Fig. 3G), binned by their intersomatic distances to the target, and compared the magnitude of influence in each spatial bin between AC and PPC. The magnitude of positive weights was higher in PPC than that in AC for spatial bins 50 and 100 µm from the target (Fig. 3F; 50 µm p = 0.0001; 100 µm p = 0.0001; 150 µm p = 0.4323; 200 µm p = 0.0315; 250 µm p = 0.0085; 300 µm p = 0.3221; 350 µm p = 0.2590; 400 µm p = 0.4143). When we compared the magnitudes of negative weights by distance bin, a complementary pattern emerged. At distances greater than or equal to 100 µm, the magnitude of negative influence was greater in AC than in PPC (Fig. 3G; 50 µm p = 0.5730; 100 µm p = 0.0015; 150–325 µm p = 0.0001; 350 µm p = 0.0002; 400 µm p = 0.0001). Bonferroni-corrected significance for these analyses is indicated in Figure 3F,G.

To summarize, we characterized the spatial distributions of positively and negatively weighted neurons in the population response to each target's stimulation. The results were consistent with the single neuron analyses in Figure 2, as positive weights were larger in magnitude overall in PPC than in AC, and negative weights were larger in magnitude in AC than in PPC, especially for neurons at farther distances from the targeted neuron.

Influence and behavioral state

While the mice in our study were not performing a task, they were allowed to run voluntarily on the spherical treadmill throughout the imaging sessions. In this context, mice tend to shift between low arousal states with little movement and high arousal states with running (McGinley et al., 2015; Vinck et al., 2015; Khoury et al., 2023). In AC, the relationship between neural activity and running speed is heterogeneous, with positive and negative modulation of activity across neuron types (Bigelow et al., 2019; Yavorska and Wehr, 2021; Khoury et al., 2023). In PPC, on the other hand, activity is positively modulated by running, especially in SOM neurons (Khoury et al., 2023). Given the different relationships between running behavior and activity in AC and PPC, we reasoned that the local influence of E neurons could be differentially linked to behavioral state in the two areas. Specifically, we expected that the stronger running modulation of population activity in PPC could lead to different network states with stronger or weaker local recurrence.

To test the hypothesis that running behavior modulates local recurrent excitatory interactions more strongly in PPC, we first measured the mean running speed during the 20 imaging frames (660 ms) preceding each photostimulation. We then split the distribution of prestimulus running speeds into quartiles, for each imaging session separately. The first quartile corresponded to mice being stationary on the treadmill. We then measured the mean influence of E neuron stimulation on other neurons in each running speed quartile, restricting this analysis to include only neurons that were significantly positively influenced when considering all trials and running speeds. The magnitude of positive influence on Non-SOM neurons was not affected by running speed in AC (influence in running speed quartiles 1 vs 2: p = 0.4439; 2 vs 3: p = 0.8644; 3 vs 4: p = 0.4730; 1 vs 4: 0.2342; Fig. 4A). However, the magnitude of positive influence on Non-SOM in PPC decreased systematically with running speed (influence in running speed quartiles 1 vs 2, 2 vs 3, 3 vs 4 and 1 vs 4: p < 0.0001; Fig. 4A). To examine the full distribution of runn