Rats were overdosed with sodium pentobarbital (100mg/kg body weight), and perfusion fixed using 2.5% paraformaldehyde and 2.5% glutaraldehyde in a 0.1M phosphate buffer (PB; pH 7.4). Brain sections (100–300μm thickness) were prepared using a microslicer (Dosaka EM, Kyoto, Japan). After the sections were processed using the rapid Golgi technique as described elsewhere [11, 12], the tissues were dehydrated and flat embedded in epoxy resin and preliminarily examined using a light microscope [8]. Appropriately Golgi-impregnated neurons in the hippocampus were selected for the following procedures. Thick sections of selected neurons were remounted for further sectioning. Two to five-micrometer thick sections were cut using an ultramicrotome (Ultracut, Reihert-Jung, Wetzlar, Germany) and collected on copper grids. Each specimen was first photographed using a light microscope for reference purposes, then examined using UHVEM (Hitachi H-1250M; National Institute for Physiological Sciences, Okazaki, Japan) at an accelerating voltage of 1,000kV. Stereopaired photomicrographs were prepared using UHVEM by tilting the specimen stage ±8°.

Rats were deeply anesthetized using intraperitoneal injections of 50mg/kg body weight sodium pentobarbital and bilaterally injected with 1μL of 0.2% cholera toxin β subunit-horseradish peroxidase conjugate (CTb-HRP; List Laboratories, Cupertino, CA, USA) into the bulbocavernosus muscles. Rats were overdosed with sodium pentobarbital (100mg/kg body weight) 48–96h after the CTb-HRP injection, and then perfusion fixed with 4% paraformaldehyde, 0.2% glutaraldehyde, and 1.25% picric acid in a 0.1M PB solution (pH 7.4). Spinal cords were immediately removed and immersed in the same fresh fixative for 3h. Spinal sections (L5-L6 level; 30μm thickness) were prepared using a microslicer (Dosaka EM). Next to visualize SNB somata and dendrites, retrogradely labeled with CTb-HRP, the tetramethylbenzidine (TMB)/diaminobenzidine- (DAB-) nickel method was performed as previously described [32]. Sodium tungstate was used as the stabilizer. Sections were then placed in a 0.1M PB with 25% sucrose and 10% glycerol for 1h for cryoprotection, then freeze-thawed using liquid nitrogen to enhance the penetration of the antibodies. After blocking with phosphate buffered saline (PBS; pH 7.4) containing 0.05% Triton X-100, 1% normal goat serum, and 1% BSA for 2h, the sections were incubated in rabbit anti-GRP serum (Phoenix Pharmaceuticals, Burlingame, CA, USA) at a 1:5,000 dilution in the blocking solution, for 5 days at 4°C. After washing with PBS, the sections were treated with biotinylated goat anti-rabbit IgG (Nichirei, Tokyo, Japan), at a 1:1,000 dilution in the blocking solution for 3h at room temperature. After washing with PBS, GRP immunoreactivity was developed using the streptavidin-biotin-HRP complex/DAB-nickel method, as previously described [29, 33]. After washing with 0.1M PB, the sections were placed in a 0.1M PB solution with 1% OsO₄ for 90min, dehydrated, and flat embedded in epoxy resin. The embedded sections were viewed using an Olympus Optical BH-2 microscope (Tokyo, Japan), and regions that may contain GRP synapses were selected. These sections were further cut into serial 2μm-thick semithin sections and collected on copper grids coated with collodion film. Each specimen was first photographed using a light microscope for reference purpose, then examined using UHVEM at an accelerating voltage of 1,000kV and conventional transmission electron microscopy (EM; JEM-1220, JEOL, Tokyo, Japan) at an accelerating voltage of 80kV. Stereopaired photomicrographs using UHVEM were prepared by tilting the specimen stage ±8°. Neuronal profiles were identified using the criteria described by Peters et al. [34].

Retrograde tracing is a reliable neuroanatomical method used to locate the somata and dendrites of the motoneurons that innervate the somatic muscles [29, 32, 35]. Neurons, retrogradely labeled with CTb-HRP, were visualized by the TMB method, and were observed only in the SNB at the fifth and sixth lumbar segments of the spinal cord. Using light microscopy, we found densely stained somata and dendrites that were diffusely distributed, or in a dot-like fashion, and often appeared to make close appositions with immunolabeled GRP-immunoreactive axons. These appositions were particularly abundant in the dorsal gray commissure (DGC). GRP-immunoreactive somata were not observed in the SNB.

In terms of their ultrastructures, the TMB reaction products in the CTb-HRP-labeled SNB neurons were also diffusely distributed within the cytoplasm, as revealed by UHVEM and conventional EM [8, 32]. The TMB reaction products were more electron-dense than the DAB reaction products of the GRP-immunoreactive axons [8]. The TMB reaction products were electron-dense small bodies, with a 3D radial spindle-like shape in clusters of various sizes within the cytoplasm and dendrites (Figure 2). The DAB reaction products were visualized as fine, fuzzy materials that were homogeneously distributed throughout the axons and terminals of the GRP neurons (Figure 2). Strikingly, the 3D TMB crystalline structures were clearly observed by UHVEM and were easily distinguishable from the DAB reaction products (Figure 2). Interestingly, 3D analysis with a polarizing lens revealed that some populations of the GRP-immunoreactive terminals possibly formed synaptic contacts with the SNB motoneurons [8]. GRP-immunoreactive axons were also found to wind around a single distal dendrite within the DGC. Taken together, these results suggest that GRP-containing afferents to SNB motoneurons regulate male sexual reflexes via these synapses, since the contraction of bulbocavernosus muscles is involved in penile erection [36].