In this study, immunohistochemistry for neuronal nitric oxide synthase (bNOS-IR), nicotinamide adenine dinucleotide phosphate diaphorase histochemistry (NADPHd) and nitric oxide synthase radioassay were used to study the occurrence, number and distribution pattern of nitric oxide synthesizing neurons in the lumbar (L1-L7) and sacral (S1-S3) dorsal root ganglia of the dog. Nitric oxide synthase immunolabelling was present in a large number of small- (area <1,000 microm(2)) and medium-sized (area 1,000-2,000 microm(2)) as well as in a limited number of large-sized (area >2000 microm(2)) neurons. Although neuronal nitric oxide synthase immunolabelling and histochemical staining provided intense staining of multiple small- and medium-sized neurons in all lumbar and sacral dorsal root ganglia, immuno-labelled or histochemically stained somata exhibited little topographic distribution in individual dorsal root ganglia. Great heterogeneity was noticed in the immunolabelling of medium-sized nitric oxide synthase immunopositive neurons ranging from lightly immuno-labelled somata to heavily immunoreactive ones with completely obscured nuclei. Both staining procedures proved to be highly effective in visualizing intraganglionic fibers of various diameters. In general, the largest fibers revealed at the peripheral end of lumbar and sacral dorsal root ganglia were larger, 6.49-9.35 mum in diameter, while those running centrally and proceeding into the dorsal roots were about 30% reduced, ranging between 5.32 and 8.67 microm in diameter. Peripherally, the occurrence of nitric oxide synthase detected in axonal profiles, and confirmed histochemically, in the specimens of the femoral and sciatic nerves, is the first indication of the presence of nitric oxide synthase in the peripheral processes of somata located in L4-S2 dorsal root ganglia. Large and thin central nitric oxide synthase immunoreactive processes of L1-S3 dorsal root ganglion neurons segregate shortly before entering the spinal cord, the former making a massive medial bundle in the dorsal root accompanied by a slim lateral bundle penetrating Lissauer's tract. Quantitative assessment of the distribution of bNOS-IR and/or NADPHd-stained neurons showed a peculiar pattern in relation to spinal levels. Apparent incongruity was found in the total number of NADPHd-stained versus bNOS-IR neurons, demonstrating a clear prevalence of small bNOS-IR somata in all lumbar ganglia, while medium-sized NADPHd-stained somata clearly prevailed all along the rostrocaudal axis with a peak in L5 ganglion. While the number of small bNOS-IR neurons clearly outnumbered NADPHd-stained and NADPHd-unstained somata in S1-S3 ganglia, an inverse relation appeared comparing the total number of medium-sized NADPHd-stained and NADPHd-unstained somata compared with the number of moderate and intense bNOS-IR neurons. Densitometry of bNOS-IR and NADPHd-stained neurons in lumbar and sacral ganglia revealed two distinct subsets of densitometric profiles, one relating to more often found medium-sized bNOS immuno-labelled and the other, characteristic for moderately bNOS immunoreactive somata of the same cell size. Considerable differences in catalytic nitric oxide synthase activity, determined by conversion of [(3)H]arginine to [(3)H]citrulline were obtained in lumbosacral dorsal root ganglia all along the lumbosacral intumescence, the lowest (0.898+/- 0.2 dpm/min/microg protein) being in the L4 dorsal root ganglion and the highest (4.194+/-0.2 dpm/min/microg protein) in the S2 dorsal root ganglion.