Eight adult Fischer 344 rats (4 male, 4 female; Harlan Sprague Dawley) were anesthetized with 2 ml/kg of a 25 mg/ml ketamine, 1.3 mg/ml xylazine, and 0.25 mg/ml acepromazine cocktail, and injected with 5.0 μl colchicine (10 μg/μl in aCSF) at the cortical locations listed below at a rate of 0.5 μl/min using a 10 μl Hamilton syringe. Following infusion, the needle remained in place for four minutes to allow adequate diffusion to the surrounding tissue. Four animals received bilateral injections in the primary motor cortex (M1) at +1.2 mm anterior (A/P) and ±2.5 mm lateral (M/L) to bregma. Half the solution was injected at 1.6 mm below the surface of the brain (D/V), and the rest at 1.0 mm. The remaining four animals all received unilateral injections in the prefrontal cortex (A/P: +3.0, M/L: +0.5, D/V: -2.3 and -1.8), temporal cortex (A/P: -6.5, M/L: +5.0, D/V: -7.5 and -7.0), and parietal cortex (A/P: -3.0, M/L: +5.0, D/V: -1.8 and -1.3). After 48 h, animals were deeply anesthetized and transcardially perfused with 250 ml cold phosphate buffered saline (pH 7.4), followed by 250 ml of cold 2% paraformaldehyde + 0.2% parabenzoquinone in 0.1 M phosphate buffer. Brains were extracted, postfixed for 2 h in the same fixative, and cryoprotected in 0.1 M phosphate buffer containing 30% sucrose for at least 72 h at 4°C. Coronal sections (40 μm) were cut on a freezing sliding microtome and stored in cryoprotectant (TCS) at 4°C until further processed for immunohistochemistry.

Sequential double-label immunohistochemistry was used to visualize neurons expressing NGF and either GABAergic or glutamatergic cell markers. Free-floating sections were washed in Tris Buffered Saline (TBS), permeabilized with 0.25% Triton X-100, and nonspecific labeling was then blocked with 5% donkey serum. Sections were incubated for 72 h at 4°C in rabbit anti-NGF antibody (Conner and Varon, 1992) diluted 1:1000 in TBS, 0.25% Triton X-100, and 5% donkey serum. Following primary antibody incubation, sections were incubated in donkey anti-rabbit biotin-conjugated IgG (1:200; Vector Laboratories, Burlingame, CA, USA). Tyramide signal amplification (TSA; PerkinElmer, Waltham, MA, USA) was applied to amplify the NGF signal, after which sections were washed in TBS and incubated in Alexa Fluor 488 or 594-conjugated streptavidin (Invitrogen, Carlsbad, CA, USA) diluted 1:200 for 3 h at 4°C. After a brief wash, sections were incubated in both mouse anti-glutamate decarboxylase (GAD) 65 (GAD-6, AntibodyRegistry:AB_528264,1:2000; Developmental Studies Hybridoma Bank, Iowa City, IA, USA) and mouse anti-GAD67 (AnitbodyRegistry: AB_2278725, 1:1500; Millipore, Temecula, CA, USA), or in mouse anti-parvalbumin (AntibodyRegistry: AB_2174013, 1:30000; Millipore, Temecula, CA, USA), mouse anti-calbindin-D-28K (AntibodyRegistry: AB_476894, 1:1500; Sigma-Aldrich, St. Louis, MO, USA), or mouse anti-Ca²⁺/calmodulin-dependent protein kinase IIα (CaMKIIα, AntibodyRegistry: AB_2067919, 1:1500; Millipore) for 72 h at 4°C. Finally, sections were washed, incubated in Alexa Fluor 594 or 488-conjugated donkey anti-mouse (Invitrogen) for 3 h at room temperature, washed again, mounted on glass slides, and coverslipped in Fluoromount-G (Southern Biotech, Birminghan, AL, USA). In order to maximize identification of GABAergic cells, GAD65 and GAD67 antibodies were co-incubated. A subset of sections was coverslipped in ProLong Gold with DAPI (Invitrogen) for visualization of cell nuclei.

The NGF antibody used in this study is an affinity-purified polyclonal raised in rabbit against purified mouse NGF (Conner and Varon, 1992). The antibody recognizes purified mouse and recombinant human NGF but does not cross react with recombinant BDNF or NT-3 (Conner and Varon, 1996). Furthermore, the immunoreactive pattern of NGF expression in the rat brain closely matches that obtained from in-situ analysis in the rat brain (Conner and Varon, 1997).

The monoclonal antibody GAD-65 (Developmental Studies Hybridoma Bank, Gad-6) was produced by immunizing mice with GAD protein immunoaffinity-purified from rat brain. Western blot analysis of rat brain homogenates revealed the antibody selectively recognizes GAD-65 but not GAD-67 (Chang and Gottlieb, 1988). Additional studies have shown that the GAD-6 antibody recognizes an epitope located between amino acids 475–571 of the C-terminus of GAD-65 (Butler et al., 1993).

The GAD-67 mouse monoclonal (Millipore, MAB 5406, lot: 25010139) was raised against amino acid residues 4-101 of human GAD-67, and recognizes a single 67-kDa band on Western blot analysis of rat brain (manufacturer’s technical information). Preincubation of the antibody with a GST-GAD-67 fusion protein resulted in no immunopositive signal in the brain (Ito et al., 2007).

The mouse monoclonal anti-CaMKIIα (Millipore, MAB 8699, lot: LV1366080) specifically recognizes the alpha subunit of calcium/calmodulin-dependent protein kinase II. Western blot analysis demonstrates that the antibody identifies a single band of 50 kDa and recognizes both phosphorylated and unphosphorylated forms (Erondu and Kennedy, 1985).

Monoclonal anti-calbindin-D-28k (Sigma–Aldrich, C9848, lot: 088k4799) is derived from BALB/c mice immunized with purified bovine kidney calbindin-D-28k. Immunoblotting showed the antibody recognizes a 28-kDa band, and does not to react with similar molecules, such as calbindin-D-9K, calretinin, myosin light chain, and parvalbumin (manufacturer’s technical information). Preabsorption with a calbindin-D-27 kDa protein purified from chick and rat brains was shown to eliminate calbindin immunostaining in the brain (Pasteels et al., 1987).

Anti-parvalbumin (Millipore, MAB 1572, lot: LV1378387) was collected from mice immunized against parvalbumin purified from frog muscle. The monoclonal antibody is directed against an epitope at the first Ca²⁺-binding site and immunoblot analysis demonstrates it recognizes a brain protein of 12 kDa (manufacturer’s technical information).

Images were captured using an Olympus AX70 with Magnafire software (version 2.0; Karl Storz Imaging, Goleta, CA, USA). Because NGF labeling was reduced substantially approximately 2 mm anterior/posterior to colchicine injection sites, presumably due to a lack of colchicine diffusion and the resulting absence of somatic NGF accumulation, only sections within 1.5 mm of each colchicine injection site were analyzed. Every 10^th section (400 μm) was examined within a cortical region. Cells were manually counted, and at least three sections were analyzed in each cortical region and antibody combination per subject. Only three animals were evaluated in the temporal region due to undetectable NGF labeling in the fourth. Single and double-labeled cells were quantified using both single and double-channel images. Several criteria were used to identify labeled cells, including size, morphology, signal vs. background, and coincident DAPI labeling when assessing DAPI-stained tissue. In a subset of sections, 5 μm z-stacks were collected using an Olympus Fluoview FV1000 to ensure neuronal localization of the labeled object. Due to our interest in the proportion, and not the absolute number, of double-labeled cells, stereological methods were not used.

The percentage of double-labeled cells per immunoreactive (IR) cell group was determined for each image field. Mean ± standard error was calculated for each cortical region examined. One-way analysis of variance (ANOVA) was used to evaluate differences among cortical regions. Fisher’s HSD was used for post-hoc analysis. All statistical analyses were carried out with SPSS 15.0 for Windows.

Controls included omission of primary antibodies, omission of secondary antibodies, and replacement of primary antibody with nonspecific antibody (rabbit IgG). All manipulations had the expected effects and supported the assertion that labeled cells represent true antigen labeling by their corresponding antibodies.

Distinct NGF labeling was visible within a radius of 1.5 mm from colchicine cortical injection sites. Within these areas, NGF labeling was predominantly confined to cell somata (Figure 1). Rarely, one or more cellular processes could also be distinguished. Outside of this 1.5 mm radius, NGF labeling was virtually undetectable in the cortex. As cortical expression of the NGF receptors TrkA and p75 are confined to cholinergic corticopetal fibers, the observed labeling of NGF is unlikely to reflect endocytosed NGF, but instead is indicative of NGF-producing cells (Holtzman et al., 1995; Rossi et al., 2002; Stephens et al., 2005). As reported previously (Ribak et al., 1978), colchicine treatment intensified GAD labeling in cell bodies as well as neuronal processes. Colchicine had no detectable effect on labeling for parvalbumin, calbindin, or CaMKIIα.

Nerve growth factor co-localized extensively with the GABAergic cell markers GAD65 and GAD67, regardless of the cortical area examined (Figure 1; Table 1). Overall, 91 ± 0.9% of NGF-labeled cortical cells also labeled for GAD65/67. The percentage of NGF-labeled cells co-expressing GAD65/67 showed little difference among the prefrontal (90.0 ± 1.5%), motor (91.7 ± 1.5%), parietal (89.6 ± 3.5%) and temporal (93.4 ± 8.1%) cortices (one-way ANOVA; p = 0.78). Conversely, NGF co-localized with only 55 ± 2.3% of all GAD65/67-labeled cells. To determine if NGF production was restricted to a specific subtype of GABAergic neuron, we co-labeled tissue for NGF and either parvalbumin or calbindin (Figure 2). NGF-labeled cells were observed to colocalize with both markers. However, NGF colocalization with parvalbumin (67.8 ± 3.6%) was over 2× greater than with calbindin (29.1 ± 3.9%). Additionally, NGF-IR cells constituted less than half of all parvalbumin (47.7 ± 4.6%) and calbindin (25.7 ± 4.9%) immunoreactive cells.

Nerve growth factor-expressing neurons were observed throughout all cortical layers. Previous studies have reported uneven distribution of NGF-labeled neurons in the cortical laminae (Pitts and Miller, 2000; Patz and Wahle, 2006). Quantitative analysis by layer was not performed in the current study, however, as NGF labeling intensity diminished with increasing distance from the colchicine injection site.

Labeling for CaMKIIα was primarily observed within cell somata and proximal processes (Figure 3). Unlike the extensive co-localization seen with NGF and GABAergic markers, NGF-labeled cells rarely co-localized with CaMKIIα-labeled cells (Figure 3; Table 2). In total, 4.9 ± 1.1% of NGF-immunoreactive cells were co-labeled with CaMKIIα antibodies. Co-localization differed significantly by cortical region (one-way ANOVA; p = 0.03); Fisher’s post hoc revealed that the prefrontal cortex had a greater proportion of double-labeled NGF cells (7.6 ± 2.1%) compared to the primary motor cortex (2.4 ± 1.0%; p = 0.02) and parietal cortex (2.9% ± 1.5; p = 0.01).

Cells immunoreactive for CaMKIIα greatly outnumbered those labeled by NGF antibodies. The overall proportion of CaMKIIα-labeled cells simultaneously expressing NGF signal was 2 ± 0.6%. This percentage differed significantly by region (one-way ANOVA p = 0.003), with the prefrontal cortex exhibiting a greater proportion of double-labeled NGF/CaMKIIα cells (3.7 ± 1.1%) than the primary motor cortex (0.8% ± 0.3; p = 0.001), the parietal cortex (1.4% ± 0.6; p = 0.01), and the temporal cortex (1.5 ± 0.6%; p = 0.01).