Activity correlation imaging: visualizing function and structure of neuronal populations

Abstract

For the analysis of neuronal networks it is an important yet unresolved task to relate the neurons' activities to their morphology. Here we introduce activity correlation imaging to simultaneously visualize the activity and morphology of populations of neurons. To this end we first stain the network's neurons using a membrane-permeable [Ca(2+)] indicator (e.g., Fluo-4/AM) and record their activities. We then exploit the recorded temporal activity patterns as a means of intrinsic contrast to visualize individual neurons' dendritic morphology. The result is a high-contrast, multicolor visualization of the neuronal network. Taking the Xenopus olfactory bulb as an example we show the activities of the mitral/tufted cells of the olfactory bulb as well as their projections into the olfactory glomeruli. This method, yielding both functional and structural information of neuronal populations, will open up unprecedented possibilities for the investigation of neuronal networks.

Figure 1

Principle of ACI. (a) A fluorescence image of the Xenopus OB stained with [Ca²⁺] indicator Fura-2. The processes of the neurons are hardly visible due to a lack of contrast. (b) A pixel-based map of cross correlation values (Xcorr map) calculated with respect to the signal of the marked soma (the reference signal). In this map, the brightness of individual pixels encodes the degree of cross correlation between the pixels' fluorescence signals and the reference signal. Three processes of the cell can be visualized in this map because of their correlated [Ca²⁺] signals (right, the traces represent averaged signals of the bracketed areas). (c) Maps of correlation values for four neurons (marked in the raw Fura-2 image, left), coded with different colors (middle), and overlayed on the raw fluorescence image (right). The [Ca²⁺] signals used to generate the maps are shown on the corresponding correlation maps (middle). Scale bar, 20 μm.

Figure 2

Microscopic setup. (a) Schematics of optical design. Laser light (488 nm) issuing from a glass fiber is collimated and filtered. A cylindrical lens focuses the Gaussian beam onto a line on the scan mirror. An image of this line is projected into the backfocal plane of the objective, resulting in a perpendicular line in the object plane. Moving this line in y-direction scans the object in 2D. Scans in 3D (inset) can be achieved by additional objective positioning (z) using a piezo actuator. The light emitted from the sample is gathered by the objective's aperture. It passes the tube lens, the scan lens, the dichroic mirror, and emission filter, and is then imaged by a single lens onto a linear CCD array. Note the absence of a slit aperture in front of the camera. Colored arrows indicate the propagation for excitation (green) and emission (red) light. (b) Schematics of light propagation in x- and y-direction for excitation (green lines) and emission (red lines). Glass fiber, collimator, and filters are omitted for clarity. In x-direction (light colors, broken lines), the setup is identical to a standard confocal microscope, the emission pinhole being replaced by the width of the CCD line detector. In y-direction (dark colors, solid lines), a cylindrical lens introduces a focus on the scan mirror and subsequently in the back-focal plane of the objective, comparably to a wide-field illumination setup, resulting in a focal line in the front focal plane (object plane) of the objective. By tilting the scan mirror, the line is moved across the specimen (not depicted). The emission pathway of the y-direction is similar to the x-direction, with the exception that the light emerging from the points along the line are not descanned (solid red line), thus forming an image of the line on the CCD array. The z-position (position of the objective) is controlled by a piezo-driven actuator. GF, glass fiber; SM, scan mirror; DM, dichroic mirror; CL, cylindrical lens; SL, scan lens; TL, tube lens; PA, piezo actuator; O, objective; S, specimen (object plane); CCD, ccd camera.

Figure 3

ACI together with fast confocal 3D scanning reveals network function and structure. (a) Mean projections over time of raw intensity images for three different z-positions as indicated. Superimposed are a region of interest (arrowhead) placed on a glomerulus and its corresponding time trace, taken as reference trace. Vertical scale bar 10% of ΔF/F₀, horizontal scale bar refers to time trace (10 s). For length scale see b. (b) By applying the proposed correlation analysis to time series of image stacks, we obtained stacks of correlation maps as those presented in Figure 1. Shown are 3 out of 18 correlation maps at the z-positions indicated in a. The maps are based on correlation coefficients calculated with respect to the reference time trace. The colorbar codes the correlation values of the map. Scale bar, 10 μm, applies also to a. (c) Maximum z-projection reveals the morphology of the neurons belonging to this functional unit. Due to the sparseness of the functional labeling, the processes are clearly identifiable and the connectivity between somata and glomerulus is obvious. Same colormap as b. (d) By choosing reference traces from other ROIs (same experiment), multiple correlation maps were generated. Shown are six examples (each being a maximum z-projection like the one in d). All maps are scaled to the same interval used in b as indicated by the colorbar. (e) Left, [Ca²⁺] signals of 190 different ROIs, arranged as a matrix with time in x- and ROI index in y- direction. Color code of the time traces, dark blue to light red for low to high [Ca²⁺], respectively. Right, four different color lookup tables chosen to highlight different features of the network. (f–i) Combining the correlation maps of different ROIs using the corresponding color lookup tables in e. In these maps, the hue of each pixel is determined by the hue of the ROI that most correlate to the pixel's signal. The intensity was determined by the degree of correlation.

Figure 4

ACI in 3D using a conventional confocal laser scanning microscope. (a) Correlation maps of one neuron at different z-positions reconstructed by sequentially acquiring time series of confocal images at different z-planes. An overlap between optical slices allows the selection of ROIs (arrows) belonging to the same cell at different z-planes. The reference signals used to generate the corresponding correlation maps are superimposed. (b) Maximum z-projection of the correlation maps of this cell. (c) Multicolor visualization of the network structure resulting from the superposition of the correlation maps of different neurons.

Figure 5

ACI and dye injection. (a) Intensity image after dye injection with a patch pipette (Alexa 532). The neuron's processes that are visible in the correlation map b are all present here. Some dendrites appear only very faint, indicating their small diameter (inset, empty and solid arrowheads). The inset is a projection of only three out of 24 z-planes. Scale bar 10 μm. (b) From this correlation map, the soma was selected for dye injection (black star). The inset shows a magnification of the indicated part (same as in a), with the empty arrowhead pointing at a part of a dendrite that disappears into noise (solid arrowhead). The marked dendrite (green star) does not belong to the solid neurons and only seems to branch off of the primary dendrite due to the depicted projection of the image stack. (c) The overlay of the correlation map with the injection image confirms that the majority of the structures is present in both images. (d) Correlation map acquired with a higher magnification objective. Thin secondary dendrites can be identified unambiguously (solid arrowheads). Scale bar 10 μm.

Figure 6

Specificity of ACI. The probability that a pixel's signal becomes more correlated to a “wrong” neuron as a function of SNR for three values of ICC. The ability to define the identity of a pixel increases (i.e., error probability decreases) as SNR becomes larger or the correlation between neurons (ICC) becomes smaller.