Research: Imaging Studies

CNIC researchers use a combination of confocal and multi-photon microscopy to obtain high quality images. All datasets are acquired on-site through the use of the latest hardware and software techniques.

Confocal Microscopy: A Case Study

Figure 1: Maximal projection of an image stack obtained by confocal microscopy before deconvolution.

Figure 2: Maximal projection of the same image stack in Figure 1 after deconvolution has been applied. Note the reduction in optical smear along the Z axis.

Four 9 month old male mice (C57Bl/SJL) were used. Animals were anesthetized with choral hydrate (15% aqueous solution, i.p.) and were perfused transcardially with 4% paraformaldehyde and 0.125% glutaraldehyde in phosphate buffer saline (PBS; pH 7.4). The brains were then carefully removed from the skull and postfixed for 6 hours.

For intracellular injections, brains were coronally sectioned at 200 μm on a Vibratome (Leica, Nussloch, Germany). The sections were then incubated in 4,6-diamidino-2-phenylindole (DAPI; Sigma, St. Louis, MO, USA), a fluorescent nucleic acid stain, for 5 minutes, mounted on nitrocellulose filter paper and immersed in PBS.

Using DAPI as a staining guide, individual layer II/III pyramidal neurons of the frontal cortex were loaded with 5% Lucifer Yellow (Molecular Probes, Eugene, OR, USA) in distilled water under a DC current of 3-8 nA for 10 minutes, or until the dye had filled distal processes and no further loading was observed.

Tissue slices were then mounted and coverslipped in Permafluor. Dendritic segment and spine imaging was performed using a Zeiss 410 confocal laser scanning microscope (Zeiss, Thornwood, NY, USA) using a 488 nm excitation wavelength, using a 1.4 N.A. Plan-Apochromat 100x objective with a working distance of 170 μm and a 5x digital zoom. After gain and offset settings were optimized, segments were digitally imaged at 0.1 μm increments, along the optical axis. The confocal stacks were then deconvolved with AutoDeblur (MediaCybernetics, Bethesda, MD, USA).

Requirements for Spine Detection

We have found that the algorithms used by NeuronStudio can successfully detect spines at resolutions as low as 0.2 μm in any direction. For reliable shape classification, we recommend that voxel resolution be maintained at 0.1 μm or higher in all directions. Whatever the chosen resolution, our method requires that the LSM data be properly deconvolved in order to reduce the optical smearing introduced by the point spread function (PSF) of the microscope as well as to filter out any shot noise created by the CCD camera during digitization.