Animals, doxycycline administration, and stereotactic injection
All animal experiments were done according to the guidelines of and approved by the veterinary office of the Canton of Zürich, Switzerland. NIPP1*-GFP rtTA2 have been described before. Doxycycline (DOX) (Grovet, Amsterdam) was administered as wet food mixture. 600 mg doxycycline was mixed with 500 g wet food (equals to 100 g dry food) and 20 g wet food with DOX was given per mouse, freshly prepared every day. Mice were stereotactically injected with 1 µl of retroviral suspension, as described before. The coordinates were 2 mm posterior of the bregma, 1.5 mm lateral of the midline, and 2.3 mm ventral from the skull, with bregma and lamda leveled and measures taken at the bregma. Injections were performed using a stereotactic frame (Kopf) and a Hamilton syringe (Hamilton). Animals were anesthetized using Ketamin (Ratiopharm) and Xylazin (Bayer). For BrdU experiments, we used DOX-tg-NIPP1* (n = 3) and non-DOX-tg-NIPP1* (n = 5); for IdU/CldU experiments we used DOX-tg-NIPP1* (n = 7), non-DOX-tg-NIPP1* (n = 4), single transgenic DOX-TetO-NIPP1* (n = 9), single transgenic non-DOX-TetO-NIPP1* (n = 12); for retroviral experiments: DOX-tg-NIPP1* (n = 4) and non-DOX-tg-NIPP1* (n = 4); for EM experiments DOX-tg-NIPP1* (n = 1) and non-DOX-tg-NIPP1* (n = 1). Thymidine analogues BrdU, CldU or IdU (Sigma) were administered as intra peritoneal (i.p.) injections in doses equimolar to 100 mg/kg of BrdU at the noted timepoints. For antigen retrieval, sections were incubated in 2 N HCl at 37°C for 30 min, followed by a wash in 0.1 M Borate buffer. CldU and IdU were detected by antibodies directed against BrdU, one of which (rat α-BrdU, Abcam) recognizes, next to BrdU, also CldU but not IdU; the second one (mouse α-BrdU, BD) recognizes all three thymidine analogues, but the affinity to CldU is weaker than to the other two and it can be washed off with Tris-HCl (40 mM, pH 8.0, 0.5 M NaCl, 1% Tween). That way, if the mouse α-BrdU is first applied to the sections and washed off from the CldU afterwards, followed by a staining with the rat α-BrdU, each antibody specifically detects one of the thymidine analogues.
Mice were perfused transcardially with 0.9% NaCl, followed by freshly prepared chilled 4% formaldehyde solution (pH 7.4). The brains were postfixed overnight on 4% PFA in 4°C on a shaker and transferred to 30% sucrose solution on the next day for cryoprotection. For sectioning, the brains were cut to 40-µm-thick coronal sections on a freezing sliding microtome (Microm). Sections were collected to a cryoprotectant solution (CPS, 25% Ethyleneglycol, 25% Glycerin, 0.1 M PO4) in a series of 12. For quantitative analysis, two series of each brain were stained and analyzed. For immunohistochemical staining, sections were washed and blocked in 3% donkey serum (Millipore) and 0.2% Triton X100 (Sigma). Primary antibodies used were goat α-DCX (Santa Cruz), rat anti-BrdU (Abcam), mouse α-BrdU (BD), mouse α-NeuN (Millipore), and goat α-Prox1 (Santa Cruz). Secondary antibodies were donkey α-primary antibody species, conjugated to different fluorophores (Jackson ImmunoResearch). For determining the number of thymidine analog-positive cells, the cells were counted on all sections of two series and the number multiplied by 6. This number was then denoted as the final number of cells per DG.
Plasmids and viruses
Retroviral vectors based on a murine Moloney leukemia virus expressing green fluorescent protein (GFP) under the control of the chicken β-actin (CAG) promoter were produced (CAG-GFP) and injected as described previously. Microscopic counting of thymidine analog-positive cells was done with a 40x objective on an Zeiss Axiovert Observer-D1 inverted microscope. For measuring the dendritic length, z-stacks were acquired using a 20× objective on the Leica SP2 AOBS laser scanning confocal microscope. A maximum projection of the stacks was measured using ImageJ (NIH) with the NeuronJ plugin for analysis of dendritic length and branch points. For measuring spine density, z-stacks were acquired using the Leica SP2 AOBS with the 63× oil objective and a zoom of 5, a resolution of 1024×1024 pixels, and a z-step size of 0.2 µm. Those stacks were then deconvoluted using Huygens essential deconvolution software (SVI). Dendrites and spines were tracked in semi-3D using NeuronStudio, and the spines were categorized according to their morphology. The data from NeuronStudio was evaluated using a VBS MS Excel Tool for extraction and automated analysis on the collected data for spine numbers, densities, and classifications. Images were processed using Photoshop CS5 (Adobe), ImageJ (NIH), FIJI or Imaris (Bitplane). To estimate the number of spines per cell, the density was multiplied with the dendritic length. It has to be considered, however, that spines are not uniformly distributed on the dendrites. As virtually no spines are formed in the granule cell layer (GCL), we subtracted 60 µm from the average length to exclude the portion of the GCL of the dendrites. The dendritic portion of the GCL is not included in the spine density numbers, as we did not acquire images there. Therefore, we multiplied the measured number of spines/µm with the average measured dendritic length reduced by 60 µm to obtain the estimated number of spines per cell.
For electron microscopy of synaptic spines, mice were perfused transcardially with 0.9% NaCl and 4% cold PFA, 0.1 M phosphate buffer and post fixed overnight with the brain still in the skull on 4°C, followed by 48 h postfixation of the brain in 4% PFA. The brains were sectioned using a Leica vibratome to 50-µm-thick sections. Cells containing GFP were visualized using a rabbit anti-GFP (Chemicon) primary and a biotinylated anti-rabbit secondary antibody (Jackson), followed by a DAB staining (Vectorlabs Kit). Afterwards, lipidous structures were visualized by incubation in 1% OsO4. The sections were then embedded in epon resin for electron microscopy. The tissue section was glued on a resin block. Using a wide-field light microscope, stained dendrites in the embedded section were identified and landmarks created in relation to them on the surface of the section using a syringe needle. The resin block was shortened to about 4-5 mm and mounted with superconductive carbon cement (Leit-C, Neubauer Chemikalien) onto SEM- stubs. The samples were sputter-coated with 5 nm platinum (BalTEC sputter coater: MED 010), and the surface imaged in a scanning electron microscope (SEM) (Zeiss LEO 1230, SE mode). SEM surface images and the light microscopic images were combined and aligned in Adobe Photoshop CS5. The area of interest was relocated in FIB/SEM and covered with up to 1 µm carbon deposition (GIS, FIB/SEM FEI Helios600i). The surface was opened and a perpendicular section plane of about 50 × 50 µm along the stained neuron was generated by focused ion beam milling (gallium ions at 30 kV, 21 nA, and 9 nA) (Helios Nanolab 600i and Zeiss NVision 40). The surface was polished and sequentially cut with a focused ion beam at 30 kV and 2.5 nA. Between the serial sections, SEM images were taken at 2 kV and 0.17 nA at selected regions within the section plane (back-scattered electron signal of the in-lens detector at a dwell time of 30 µs per pixel using the G3-slice & view software on FEI Helios600i). Each image area was captured with an image pixel size of 5 nm. Images were taken every 30 nm, resulting in a data voxel size of 5 × 5 × 30 nm. Serial images were aligned and processed in FIJI. Structures were registered manually using either FIJI TRAK EM or Imaris (Bitplane). From the tracked contours, three-dimensional reconstructions were done using the same programs. Synapses were identified and categorized as single or multiple synapse boutons.
Statistical analysis was performed using SPSS 18 (IBM). Differences were considered significant at p <0.05. All the tests used to compare the average number of cells, spines, or dendritic length in the control vs. the Dox-fed double transgenic mice were 2-tailed independent sample t-tests. Significance indicates the p-value was below 0.05; if above it is reported as non-significant (n.s.),± represents the standard error of the mean.