The functional microarchitecture of the mouse barrel cortex

Abstract

Cortical maps, consisting of orderly arrangements of functional columns, are a hallmark of the organization of the cerebral cortex. However, the microorganization of cortical maps at the level of single neurons is not known, mainly because of the limitations of available mapping techniques. Here, we used bulk loading of Ca(2+) indicators combined with two-photon microscopy to image the activity of multiple single neurons in layer (L) 2/3 of the mouse barrel cortex in vivo. We developed methods that reliably detect single action potentials in approximately half of the imaged neurons in L2/3. This allowed us to measure the spiking probability following whisker deflection and thus map the whisker selectivity for multiple neurons with known spatial relationships. At the level of neuronal populations, the whisker map varied smoothly across the surface of the cortex, within and between the barrels. However, the whisker selectivity of individual neurons recorded simultaneously differed greatly, even for nearest neighbors. Trial-to-trial correlations between pairs of neurons were high over distances spanning multiple cortical columns. Our data suggest that the response properties of individual neurons are shaped by highly specific subcolumnar circuits and the momentary intrinsic state of the neocortex.

Figure 1. Loading Populations of L2/3 Neurons with Ca Indicators In Vivo

(A) Left: cortical region stained with Fluo-4 AM (green fluorescence). Right: the distribution of Alexa 594 during loading (red fluorescence). The dark band is a shadow cast by a blood vessel in the imaging area. (B) Higher magnification of cortical cells. Left: Fluo-4 AM image. Right: Alexa 594 image. Note the clear correspondence between the labeled cells in the Fluo-4 AM image and the unlabeled dark cells in the Alexa 594 image (right). Other dark structures in the Alexa 594 image correspond to blood vessels viewed in cross-section. (C) Cells stained with Fluo-4 AM (left) and Sulforhodamine 101 (SR 101) (right). The cells with high Fluo-4 AM fluorescence were also labeled with Sulforhodamine 101, indicating that they were astrocytes.

Figure 2. Spontaneous and Whisker Stimulation-Evoked Fluorescence Transients in the Somata of Cortical Neurons

(A) Images of spontaneous fluorescence transients. The left image shows three regions of interest corresponding to three cells over 3 s. The electrophysiological recording was performed from Cell 1. The subsequent two images show the fluorescence before and after onset of the transient in Cell 1, averaged over the time periods indicated by horizontal bars in (C) and (D). The third image shows the difference image. (B) Different 3-s trial, same region as in (A). (C) Fluorescence changes corresponding to (A). The tick mark indicates an AP recorded from Cell 1 using a cell-attached electrode. See also Video S1. (D) Fluorescence changes corresponding to (B). In this trial, Cell 1 did not produce an AP. (E) Averaged fluorescence transients following one and two spikes (Cell 1 in [A–D]). The ratio of the amplitudes is a factor of two. (F) The peak fluorescence change as a function of the number of APs (18 cells). (G and H) Examples of whisker stimulation-evoked fluorescence transients. Cell-attached recordings were made from Cell 1. In (G), only Cell 2 produced a fluorescence transient. In (H), both cells produced a transient. Note the diffuse neuropil signal (arrows). (I and J) Fluorescence changes corresponding to (G) and (H), respectively. Whisker stimuli are indicated on the bottom. (K) Fluorescence traces from different trials were overlaid. Top, traces corresponding to Cell 1 (G–J). Trials producing whisker stimulation-evoked transients (successes) were easily distinguished from trials without transients (failures). Middle and bottom, traces from other cells. In these cells, the separation between successes and failures was less clear.

Figure 3. Distinguishing Successes and Failures in Response to Whisker Deflection

(A) Left: multiple fluorescence transients from a single cell aligned on the whisker stimulus. Right: analysis of two representative trials (green and pink traces on the left). For each trace, two values, the difference F _d (solid arrows) and the amplitude A _F (dashed arrows) were computed. The dotted lines are the results of template matching. (B) Left: the F _d and A _F were plotted for each trial. Right: these points were clustered into two groups (red, successes; blue, failures). The 95% confidence ellipses are overlaid on the graph. The green and pink dots are from the two representative trials shown as the green and pink traces in (A), respectively. (C–E) Analysis for three different cells. (C) Well-separated cell. (D) Marginally separated cell. (E) Example of a cell in which successes and failures overlapped.

Figure 4. Validation of the Analysis Algorithms with Cell-Attached Recordings

(A) The distribution of the overlap between the two ellipses for experiments with cell-attached recordings (n = 34 neurons). For some cells, the number of APs was too little to allow analysis (not defined). (B) For cells with zero overlap, the correspondence between scoring successes and failures from imaging experiments and APs was very high (left). For cells with overlap, false-positive and false-negative trials were often detected (graph on right). Note the differences in y-axes between the two graphs. (C) The location of the focal plane within the soma affects detection fidelity. When the focal plane was centered on the soma (left), successes and failures were clearly distinguishable. In the same cell, when the focal plane was closer to the upper edge of the cell (right), the two clusters were less segregated. (D) The overlap increases, and detection fidelity decreases, as the focal plane is moved from the center of the soma, closer to the edge.

Figure 5. The Response Properties of L2/3 Neurons

(A) The distribution of success probabilities in response to PW and SW stimulation (n = 292 neurons). (B) The distribution of the SI, defined as the relative response probability for the PW and SW (SI: possible range, −1,1).

Figure 6. Somatotopy at the Level of Single Cells

(A) Left: the location of the imaged area in CO-stained barrels. Right: Fluo-4 AM image showing the locations of the analyzed cells. The imaged area was centered on the C3 barrel, with the C4 barrel to the right. (B) The SI of neurons in (A) as a function of their location. Only cells without overlap (Φ = 0) were included in the analysis. (C) Spatial changes in the selectivity index across 33 experiments. To overlay different experiments the SI and location were shifted so that the center of gravity of the data points was on the origin.

Figure 7. Heterogeneity of Response Selectivity

(A) Left: the location of the imaged area in CO-stained barrels. Right: Fluo-4 AM image showing the locations of the analyzed cells. (B) The response pattern of two neighboring neurons (Cell 3 and Cell 22) to stimulation of whiskers C3 (the PW) and C4. (C) The response probability of Cell 3 and Cell 22 to C3 stimulation (black) and C4 stimulation (white). (D) The difference in SI for pairs of neurons as a function of the distance between the neurons. The distance was calculated as the projection of the position vector connecting pairs of neurons onto the line that connects the two neighboring barrels. Essentially indistinguishable data were obtained using the absolute distance (length of the position vector; see Figure S3). Red circles indicate pairs of neurons whose SIs were significantly different. RF, receptive field.

Figure 8. Trial-to-Trial Correlation of Sensory-Evoked Responses

(A) The trial-to-trial correlation between the response patterns for each pair of neurons (imaged simultaneously) to PW stimulation was plotted as a function of the distance between the two neurons. (B) The same analysis as depicted in (A), but with the paired correlations plotted from the response pattern following SW stimulation. (C) Top: raw response pattern of 11 simultaneously imaged neurons to PW stimulation. Each row represents one neuron, and each column represents one trial. The neurons were sorted based on response probability (Cell 1 responded the least, and Cell 11 responded the most). Bottom: the response patterns from above were sorted so that all responsive trials were clustered towards the bottom of the chart, regardless of cellular identity and origin of that response. The number of responsive neurons in each trial remained the same. (D) Monte Carlo simulation for the correlation coefficient between “Raw data” and “Adjusted data,” from (C). The red arrow indicates the actual value. (E) The distribution of the correlation coefficient between “Raw data” and “Adjusted data” for 23 imaging sessions in which more than five neurons responded for more than 20 trials.