Neurobiotin was allowed to diffuse for 1–4 h into the distal compartments of the cells after labeling, while the animal was still under anesthesia. The animals were then transcardially perfused with cold saline, followed by 300 ml of cold fixative (4% paraformaldehyde, 0.05% glutaraldehyde, and 15% v/v saturated picric acid in 0.1 m phosphate buffer, pH ~7.4). Brains were removed, and sections of the right hippocampus were cut at nominally 70 μm thickness in the coronal plane with a vibratome (Leica VT 1000S).
We performed immunohistochemical reactions on individual free-floating sections to establish the protein expression in Neurobiotin-labeled interneurons. For incubations, standard procedures were used as described previously (Somogyi et al., 2004 (link)). Neurobiotin was visualized with streptavidin conjugated to Alexa Fluor 488 (1:1000; Invitrogen), DyLight488 (1:500 or 1:1000; Jackson ImmunoResearch), or 7-amino-4-methylcoumarin-3-acetic acid (AMCA) (1:100; Vector Laboratories). We used commercially available secondary antibodies raised in donkey (unless indicated otherwise) against the primary antibodies of the appropriate species, conjugated to AMCA (1:100; Vector Laboratories), Alexa Fluor 488 (1:1000; Invitrogen), DyLight488 (1:500; Jackson ImmunoResearch), Cy3 (1:400; Jackson ImmunoResearch), or Cy5 (1:250; Jackson ImmunoResearch). For the visualization and evaluation of most reactions, standard epifluorescent microscopy was used, with one of three upright microscopes. Because excitation and emission spectra of Alexa Fluor 488 and Dylight488 completely overlap, the same filter sets were used to detect the two fluorophores in separate sections. The filter sets, objectives, light source, and camera used with the Leitz DMRB microscope have been described in detail previously (Ferraguti et al., 2004 (link); Somogyi et al., 2004 (link)). We also used an Olympus BX61 microscope with 20× (UPlanSApo, NA 0.75), 40× (oil immersion, UPlanFLN, NA 1.3), or 60× (oil immersion, PlanApoN, NA 1.42) objectives, an EXFO mercury vapor arc lamp (Lumen Dynamics) for epifluorescent illumination, filter sets U-MNUA2 (AMCA; 360–370 nm excitation bandpass, 400 nm dichroic mirror, 420–460 nm emission bandpass; Olympus), U-MNIBA3 (Alexa Fluor 488; 470–495 nm excitation bandpass, 505 nm dichroic mirror, 510–550 nm emission bandpass; Olympus), U-MNIGA3 (Cy3; 540–550 nm excitation bandpass, 570 nm dichroic mirror, 575–625 nm emission bandpass; Olympus), and U-M41008 (Cy5; 590–645 nm excitation bandpass, 660 nm dichroic mirror, 670–730 nm emission bandpass; Chroma Technology) to separate the fluorescent light of different channels, and an Olympus XM10 monochrome cooled CCD camera controlled by CellF software (version 3.3; Olympus) to capture images. The third microscope (Carl Zeiss AxioImager.Z1) was used for standard epifluorescent imaging, for optical sectioning with a structured illumination system (Carl Zeiss ApoTome), and for confocal laser scanning microscopy (Carl Zeiss LSM 710); it was equipped with 10× (EC Plan Neofluar, NA 0.3), 20× (PlanApochromat, NA 0.8), 40× oil-immersion (PlanApochromat, NA 1.3), and 63× oil-immersion (PlanApochromat, NA 1.4) objectives. For standard epifluorescent imaging, the light source was a software-switchable light-emitting diode set (Colibri; Carl Zeiss) with 365 nm (for AMCA), 470 nm (for Alexa Fluor 488), 530 nm (for Cy3), and 625 nm (for Cy5) diodes or a wide-band mercury vapor lamp (Xcite; EXFO). For structured illumination, the mercury vapor lamp was used for all fluorophores. Although with light-emitting diodes no excitation filters are needed, because of the hardware configuration we used the same filter sets with diodes and the lamp. Images were captured with an AxioCam HRm3 monochrome CCD camera (Carl Zeiss), controlled by the AxioImager software (Carl Zeiss), and we used the filter sets (Carl Zeiss) designed for AMCA (code 49; 365 nm excitation filter, 395 nm dichroic mirror, 445/50 nm emission bandpass), Alexa Fluor 488 (code 38HE; 470/40 nm excitation bandpass, 495 nm dichroic mirror, 525/50 nm emission bandpass), Cy3 (code 43HE; 550/25 nm excitation bandpass, 570 nm dichroic mirror, 605/70 nm emission bandpass), and Cy5 (code 50; 640/30 nm excitation bandpass, 660 nm dichroic mirror, 690/50 nm emission bandpass). For optical sectioning with structured illumination, we used the grids appropriate for the given objective (L1 for 10× and 20×, M for 40×, and H1 for 63×).
The axon terminals of basket cell K111c were tested by immunofluorescence using confocal scanning microscopic detection of antibodies. Because the axon was weakly labeled by Neurobiotin, it could not be reliably detected by fluorescence, and Neurobiotin in the terminals was detected by the horse-radish peroxidase (HRP) enzyme reaction with nickel-intensified 3,3-diaminobenzidine-4 HCl (DAB) as chromogen. Fluorescence images of the same area were then matched to test for the presence of vasoactive intestinal polypeptide (VIP) and vesicular glutamate transporter type 3 (VGLUT3) immunoreactivity in the identified terminals of K111c. Image stacks were taken using an AxioImager.Z1 microscope (see above; 63× objective, LSM 710 scanning head, ZEN 5.0 software). The sequential scanning parameters were as follows: for Neurobiotin/CB1R (Alexa Fluor 488), laser 488 nm, filter 492–544 nm, pinhole 1.0 Airy unit (AU); for VIP (Cy3), laser 543 nm, filter 552–639 nm, pinhole 0.83 AU; and for VGLUT3 (Cy5), laser 633 nm, 637–757 nm, pinhole 0.55 AU. Pixel size was 90 nm (x, y). Pinhole sizes were adjusted to produce optical slice thickness of 700 nm in each channel; optical slices were taken at 500 nm intervals. Four scan lines were averaged. The 8-bit images were noise filtered by using a Median algorithm (kernel size: x = 3, y = 3, z = 1, kernel size channels = 1). Because the terminals of the cell were immunonegative for VIP, there was no signal in the Cy3 channel; therefore, the signals for Neurobiotin/CB1R (Alexa Fluor 488) and VGLUT3 (Cy5) were completely separated, and crosstalk between the detection channels was avoided.
No parts of images were modified in any way, and digital brightness and contrast adjustments were made on full frames. Although we applied standard procedures for all immunoreactions, as a result of unavoidable differences in some parameters between experiments (e.g., length of anesthesia before perfusion, quality of perfusion, number of electrode tracks made, time of storage in buffer, and the number of immunoreactions performed before the reaction), the results of immunofluorescent reactions showed some variability. Immunofluorescence signals in the Neurobiotin-labeled axonal, dendritic, or somatic compartments, as appropriate, were compared with neighboring immunopositive and immunonegative structures of similar type in the same focal plane. Care was taken that the exposure time, illuminating light intensity, and all other parameters were set to result in the correct dynamic range, so that even low-intensity signals were detected (seeFigs. 3A , 5A ). In most cases, after careful examination of the specimen, an all-or-none qualitative conclusion on the immunofluorescence signal associated with a particular Neurobiotin-labeled neuronal structure was possible. A Neurobiotin-labeled compartment, and as a consequence a cell, was classified “immunopositive” only if the immunofluorescence pattern was clear and its subcellular distribution was consistent with that expected, e.g., Golgi-apparatus-like for pro-CCK, nuclear for chicken ovalbumin upstream promoter-transcription factor II (COUP-TFII), or dendritic-membrane-associated for neurokinin receptor type 1 (NK1R). A cell was classified as “immunonegative” only if it could be successfully tested for immunofluorescence in the appropriate compartment [e.g., axon for CB1R and VGLUT3, soma for CCK and COUP-TFII, soma or proximal dendrite for calbindin, and soma or dendrite for metabotropic glutamate receptor 1-mα (mGluR1mα)] and if neighboring non-filled cells in the same picture frame were detected as immunopositive. The specimens in which multiple immunoreactions were performed were always carefully examined by comparing the patterns observed in the different channels. Results from immunoreactions were accepted only if it was ascertained that no cross-reactivity between any of the primary and secondary antibody combinations occurred and that a reliable separation between fluorescent imaging channels had been achieved. If any of the above criteria for evaluating are action was not met, the reaction was considered to be inconclusive.
Publications detailing the specificity of antibodies to calbindin (rabbit, 1:5000; mouse, 1:400), CB1R (rabbit and guinea pig, both 1:1000), CCK octapeptide (mouse, 1:5000), pro-CCK (rabbit, 1:500 or 1:5000), mGluR1mα (guinea pig, 1:1000), mGluR7a (rabbit, 1:500), NK1R (rabbit, 1:2000), prepro-tachykinin B (PPTB; guinea pig, 1:500), VGLUT3 (rabbit, 1:1000), and VIP [rabbit, 1:10000 (Dr. T. Görcs, Semmelweis Medical University, Budapest, Hungary); mouse, 1:50,000 (Dr. G. Ohning, University of California, Los Angeles, CA)] were given previously by Klausberger et al. (2005 (link), their supplemental Table 1) and to COUP-TFII (mouse, 1:250), muscarinic acetylcholine receptor 2 (M2R) (rat, 1:250), and VGLUT3 (guinea pig, 1:300) by Fuentealba et al. (2010 (link), their supplemental Table 1). The specificity of the guinea pig polyclonal antibody raised against synthetic VGLUT3 peptide fragment (AB5421, 1:2000; Millipore Bioscience Research Reagents) was tested by Montana et al. (2004) (link) in Western blot experiments. The specificity of the goat polyclonal anti-calretinin antibody (CG1, 1:1000; Swant) was tested by Western blot and by immunohistochemistry experiments (Schwaller et al., 1999 (link)). Specificity of other polyclonal antibodies used, which included antibodies raised against CCK precursor protein in guinea pig (1:500; gift from Dr. M. Watanabe, Hokkaido University, Hokkaido, Japan), against CCK octapeptide in rabbit (1:500; gift from Dr. M. Watanabe), against mouse mGluR1mα peptide fragment in rabbit (1:1000; mGluR1a-Rb-Af811-1; Frontier Institute Co.), against neuropeptide tyrosine (NPY) in rabbit (1:5000; code 22940; Immunostar), and against rat NK1R peptide fragment in rabbit (1:500, AB5060; Millipore Bioscience Research Reagents) was tested by the antibody provider, and these antibodies produced labeling patterns similar to that seen by several other previously characterized antibodies recognizing the same molecules. We also used a monoclonal mouse antibody against somatostatin (1:500, GTX71935; Gene-Tex) that produced labeling pattern similar to that seen with other somatostatin antibodies characterized previously.
Axonal and dendritic distributions were examined by light microscopy on sections processed with the HRP enzyme reaction for visualizing Neurobiotin using the glucose oxidase method for the generation of H2O2 and DAB only or nickel-intensified DAB as chromogens. The sections were osmium treated, dehydrated, and mounted on slides in epoxy resin. Two-dimensional reconstructions were made with a drawing tube with a 100× oil-immersion objective. Three-dimensional reconstructions were made with the Neurolucida software (version 9; MicroBrightField) using a 100× oil-immersion objective (Olympus UPlanFLN, NA 1.30).
We performed immunohistochemical reactions on individual free-floating sections to establish the protein expression in Neurobiotin-labeled interneurons. For incubations, standard procedures were used as described previously (Somogyi et al., 2004 (link)). Neurobiotin was visualized with streptavidin conjugated to Alexa Fluor 488 (1:1000; Invitrogen), DyLight488 (1:500 or 1:1000; Jackson ImmunoResearch), or 7-amino-4-methylcoumarin-3-acetic acid (AMCA) (1:100; Vector Laboratories). We used commercially available secondary antibodies raised in donkey (unless indicated otherwise) against the primary antibodies of the appropriate species, conjugated to AMCA (1:100; Vector Laboratories), Alexa Fluor 488 (1:1000; Invitrogen), DyLight488 (1:500; Jackson ImmunoResearch), Cy3 (1:400; Jackson ImmunoResearch), or Cy5 (1:250; Jackson ImmunoResearch). For the visualization and evaluation of most reactions, standard epifluorescent microscopy was used, with one of three upright microscopes. Because excitation and emission spectra of Alexa Fluor 488 and Dylight488 completely overlap, the same filter sets were used to detect the two fluorophores in separate sections. The filter sets, objectives, light source, and camera used with the Leitz DMRB microscope have been described in detail previously (Ferraguti et al., 2004 (link); Somogyi et al., 2004 (link)). We also used an Olympus BX61 microscope with 20× (UPlanSApo, NA 0.75), 40× (oil immersion, UPlanFLN, NA 1.3), or 60× (oil immersion, PlanApoN, NA 1.42) objectives, an EXFO mercury vapor arc lamp (Lumen Dynamics) for epifluorescent illumination, filter sets U-MNUA2 (AMCA; 360–370 nm excitation bandpass, 400 nm dichroic mirror, 420–460 nm emission bandpass; Olympus), U-MNIBA3 (Alexa Fluor 488; 470–495 nm excitation bandpass, 505 nm dichroic mirror, 510–550 nm emission bandpass; Olympus), U-MNIGA3 (Cy3; 540–550 nm excitation bandpass, 570 nm dichroic mirror, 575–625 nm emission bandpass; Olympus), and U-M41008 (Cy5; 590–645 nm excitation bandpass, 660 nm dichroic mirror, 670–730 nm emission bandpass; Chroma Technology) to separate the fluorescent light of different channels, and an Olympus XM10 monochrome cooled CCD camera controlled by CellF software (version 3.3; Olympus) to capture images. The third microscope (Carl Zeiss AxioImager.Z1) was used for standard epifluorescent imaging, for optical sectioning with a structured illumination system (Carl Zeiss ApoTome), and for confocal laser scanning microscopy (Carl Zeiss LSM 710); it was equipped with 10× (EC Plan Neofluar, NA 0.3), 20× (PlanApochromat, NA 0.8), 40× oil-immersion (PlanApochromat, NA 1.3), and 63× oil-immersion (PlanApochromat, NA 1.4) objectives. For standard epifluorescent imaging, the light source was a software-switchable light-emitting diode set (Colibri; Carl Zeiss) with 365 nm (for AMCA), 470 nm (for Alexa Fluor 488), 530 nm (for Cy3), and 625 nm (for Cy5) diodes or a wide-band mercury vapor lamp (Xcite; EXFO). For structured illumination, the mercury vapor lamp was used for all fluorophores. Although with light-emitting diodes no excitation filters are needed, because of the hardware configuration we used the same filter sets with diodes and the lamp. Images were captured with an AxioCam HRm3 monochrome CCD camera (Carl Zeiss), controlled by the AxioImager software (Carl Zeiss), and we used the filter sets (Carl Zeiss) designed for AMCA (code 49; 365 nm excitation filter, 395 nm dichroic mirror, 445/50 nm emission bandpass), Alexa Fluor 488 (code 38HE; 470/40 nm excitation bandpass, 495 nm dichroic mirror, 525/50 nm emission bandpass), Cy3 (code 43HE; 550/25 nm excitation bandpass, 570 nm dichroic mirror, 605/70 nm emission bandpass), and Cy5 (code 50; 640/30 nm excitation bandpass, 660 nm dichroic mirror, 690/50 nm emission bandpass). For optical sectioning with structured illumination, we used the grids appropriate for the given objective (L1 for 10× and 20×, M for 40×, and H1 for 63×).
The axon terminals of basket cell K111c were tested by immunofluorescence using confocal scanning microscopic detection of antibodies. Because the axon was weakly labeled by Neurobiotin, it could not be reliably detected by fluorescence, and Neurobiotin in the terminals was detected by the horse-radish peroxidase (HRP) enzyme reaction with nickel-intensified 3,3-diaminobenzidine-4 HCl (DAB) as chromogen. Fluorescence images of the same area were then matched to test for the presence of vasoactive intestinal polypeptide (VIP) and vesicular glutamate transporter type 3 (VGLUT3) immunoreactivity in the identified terminals of K111c. Image stacks were taken using an AxioImager.Z1 microscope (see above; 63× objective, LSM 710 scanning head, ZEN 5.0 software). The sequential scanning parameters were as follows: for Neurobiotin/CB1R (Alexa Fluor 488), laser 488 nm, filter 492–544 nm, pinhole 1.0 Airy unit (AU); for VIP (Cy3), laser 543 nm, filter 552–639 nm, pinhole 0.83 AU; and for VGLUT3 (Cy5), laser 633 nm, 637–757 nm, pinhole 0.55 AU. Pixel size was 90 nm (x, y). Pinhole sizes were adjusted to produce optical slice thickness of 700 nm in each channel; optical slices were taken at 500 nm intervals. Four scan lines were averaged. The 8-bit images were noise filtered by using a Median algorithm (kernel size: x = 3, y = 3, z = 1, kernel size channels = 1). Because the terminals of the cell were immunonegative for VIP, there was no signal in the Cy3 channel; therefore, the signals for Neurobiotin/CB1R (Alexa Fluor 488) and VGLUT3 (Cy5) were completely separated, and crosstalk between the detection channels was avoided.
No parts of images were modified in any way, and digital brightness and contrast adjustments were made on full frames. Although we applied standard procedures for all immunoreactions, as a result of unavoidable differences in some parameters between experiments (e.g., length of anesthesia before perfusion, quality of perfusion, number of electrode tracks made, time of storage in buffer, and the number of immunoreactions performed before the reaction), the results of immunofluorescent reactions showed some variability. Immunofluorescence signals in the Neurobiotin-labeled axonal, dendritic, or somatic compartments, as appropriate, were compared with neighboring immunopositive and immunonegative structures of similar type in the same focal plane. Care was taken that the exposure time, illuminating light intensity, and all other parameters were set to result in the correct dynamic range, so that even low-intensity signals were detected (see
Publications detailing the specificity of antibodies to calbindin (rabbit, 1:5000; mouse, 1:400), CB1R (rabbit and guinea pig, both 1:1000), CCK octapeptide (mouse, 1:5000), pro-CCK (rabbit, 1:500 or 1:5000), mGluR1mα (guinea pig, 1:1000), mGluR7a (rabbit, 1:500), NK1R (rabbit, 1:2000), prepro-tachykinin B (PPTB; guinea pig, 1:500), VGLUT3 (rabbit, 1:1000), and VIP [rabbit, 1:10000 (Dr. T. Görcs, Semmelweis Medical University, Budapest, Hungary); mouse, 1:50,000 (Dr. G. Ohning, University of California, Los Angeles, CA)] were given previously by Klausberger et al. (2005 (link), their supplemental Table 1) and to COUP-TFII (mouse, 1:250), muscarinic acetylcholine receptor 2 (M2R) (rat, 1:250), and VGLUT3 (guinea pig, 1:300) by Fuentealba et al. (2010 (link), their supplemental Table 1). The specificity of the guinea pig polyclonal antibody raised against synthetic VGLUT3 peptide fragment (AB5421, 1:2000; Millipore Bioscience Research Reagents) was tested by Montana et al. (2004) (link) in Western blot experiments. The specificity of the goat polyclonal anti-calretinin antibody (CG1, 1:1000; Swant) was tested by Western blot and by immunohistochemistry experiments (Schwaller et al., 1999 (link)). Specificity of other polyclonal antibodies used, which included antibodies raised against CCK precursor protein in guinea pig (1:500; gift from Dr. M. Watanabe, Hokkaido University, Hokkaido, Japan), against CCK octapeptide in rabbit (1:500; gift from Dr. M. Watanabe), against mouse mGluR1mα peptide fragment in rabbit (1:1000; mGluR1a-Rb-Af811-1; Frontier Institute Co.), against neuropeptide tyrosine (NPY) in rabbit (1:5000; code 22940; Immunostar), and against rat NK1R peptide fragment in rabbit (1:500, AB5060; Millipore Bioscience Research Reagents) was tested by the antibody provider, and these antibodies produced labeling patterns similar to that seen by several other previously characterized antibodies recognizing the same molecules. We also used a monoclonal mouse antibody against somatostatin (1:500, GTX71935; Gene-Tex) that produced labeling pattern similar to that seen with other somatostatin antibodies characterized previously.
Axonal and dendritic distributions were examined by light microscopy on sections processed with the HRP enzyme reaction for visualizing Neurobiotin using the glucose oxidase method for the generation of H2O2 and DAB only or nickel-intensified DAB as chromogens. The sections were osmium treated, dehydrated, and mounted on slides in epoxy resin. Two-dimensional reconstructions were made with a drawing tube with a 100× oil-immersion objective. Three-dimensional reconstructions were made with the Neurolucida software (version 9; MicroBrightField) using a 100× oil-immersion objective (Olympus UPlanFLN, NA 1.30).
