Quantitative Comparison of Growth-associated Protein GAP-43,Neuron-specific Enolase,and Protein Gene Product 9.5 as Neuronal Markers in Mature Human Intestine

Abstract

This study was performed to compare GAP-43, PGP 9.5, synaptophysin, and NSE as neuronal markers in the human intestine. GAP-43-immunoreactive nerve fibers were abundant in all layers of the ileum and colon. GAP-43 partially co-localized partially with every neuropeptide (VIP, substance P, galanin, enkephalin) studied. All neuropeptide-immunoreactive fibers also showed GAP-43 reactivity. By blind visual estimation, the numbers of GAP-43-immunoreactive fibers in the lamina propria were greater than those of PGP 9.5, synaptophysin, or NSE. In the muscle layer, visual estimation indicated that the density of GAP-43-immunoreactive fiber profiles was slightly greater than that of the others. The number and intensity of GAP-43-, PGP 9.5-, and NSE-immunoreactive fibers were estimated in sections of normal human colon and ileum using computerized morphometry. In the colon, the numbers of GAP-43-immunoreactive nerve profiles per unit area and their size and intensity were significantly greater than the values for PGP and NSE. A similar trend was observed in the ileum. Neuronal somata lacked or showed only weak GAP-43 immunoreactivity, variable PGP 9.5 immunoreactivity, no synaptophysin immunoreactivity, and moderate to strong NSE immunoreactivity. We conclude that GAP-43 is the superior marker of nerve fibers in the human intestine, whereas NSE is the marker of choice for neuronal somata.

THE GUT is innervated by the enteric nervous system, which is composed of the submucous and myenteric plexuses and their interconnections. The majority of nerve cell bodies are confined to the ganglia of submucous and myenteric plexuses and constitute the intrinsic component of the enteric nervous system. The extrinsic component includes the connections with the central nervous system by parasympathetic and sympathetic fibers. Changes in the density of innervation are caused by various pathological processes, such as ulcerative colitis (Keränen et al. 1995), irradiation (Hirschowitz and Rode 1991), massive small bowel resection (Vento et al. 1998), Chagas's disease (Santos-Buch 1979), achalasia (Mearin et al. 1993; Singaram and SenGupta 1996), Hirschsprung's disease (Larsson 1994), and intestinal neuronal dysplasia (Kobayashi et al. 1996). Such changes can be conveniently examined in full-thickness biopsy specimens of the intestine. Apart from immunohistochemical techniques revealing a particular neurotransmitter or neuropeptide, it is useful to estimate changes involving the gut innervation as a whole. Classical histological methods, such as the silver staining (Richardson 1960) used to reveal neurons or nerve fibers, are tedious to perform, are nonspecific, and they do not allow immunohistochemical co-localization studies.

Several antigens expressed by the neural tissue have been introduced as specific neuronal markers. Growth-associated protein (GAP-43) is expressed in conditions of embryonic growth, during axonal regeneration, and even at maturity in certain areas of the brain known to exhibit synaptic plasticity (Skene 1989; Meiri and Gordon-Weeks 1990; Benowitz and Routtenberg 1997). Recently, it has been shown that GAP-43 is expressed abundantly in the autonomic neurons and nerve fibers as well as in the enteric nervous system (Sharkey et al. 1990; Stewart et al. 1992) of the adult rat. Protein gene product (PGP) 9.5 is a cytoplasmic protein specific for neurites, neurons, and cells of the diffuse neuroendocrine system (Thompson et al. 1983; Lundberg et al. 1988). Synaptophysin is a structural component of synaptic vesicles but it has also been used as a marker of neurons (Wiedenmann et al. 1986). Neuron-specific enolase (NSE) is a common marker for neurites, neurons, and endocrine cells in the gut (Bishop et al. 1982). Both of these markers have been used in various contexts to reveal nerve cells and fibers.

The aim of this study was to evaluate GAP-43 as a general neuronal marker. Accordingly, we investigated its expression in the mature normal human ileum and colon, its co-localization with neuropeptides VIP, substance P, galanin, and enkephalin, and its usefulness in computerized morphometric analysis of the gut innervation in comparison with two other neuronal markers, PGP 9.5 and NSE.

Materials and Methods

Immunohistochemistry

Specimens of colon and ileum were taken from three patients undergoing resection of colon for treatment of neoplasia. None of the patients had bowel obstruction or other colon diseases. Whole-wall specimens were taken from the macroscopically normal margin of the resected colon or ileum segment. For comparison, specimens of labial and submandibular salivary glands were obtained from patients who underwent surgery for glandular tumors or from patients subjected to diagnostic biopsy. This study complies with Helsinki Declaration, and the permission for human studies was granted by the Ethical Committee of the Helsinki University Hospital. The specimens were immersed in 4% paraformaldehyde in PBS for 8 hr and transferred into 20% sucrose in PBS. Cryostat sections of 10 μm were cut on chromealum-gelatin-coated glass slides. For immunohistochemistry, the sections were incubated for 20 min with 5% normal rabbit serum at room temperature (RT) and overnight with the primary antiserum diluted in PBS at 4C. The dilutions were 1:500 for mouse anti-GAP-43 (Boehringer; Mannheim, Germany), 1:2000 for rabbit anti-PGP 9.5 (Affinity; Exeter, UK), 1:1000 for rabbit anti-NSE, 1:500 for rabbit anti-galanin (Chemicon; Temecula, CA), 1:500 for mouse anti-synaptophysin (Sigma; St Louis, MO), 1:500 for rabbit anti-VIP, 1:500 for rabbit anti-substance P, 1:500 for rabbit anti-enkephalin (Incstar; Stillwater, MN). The specificity of the following antibodies used has been characterized elsewhere: anti-GAP-43, clone 91E12 (Schreyer and Skene 1991), anti-NSE, batch AB951 (Thompson et al. 1983), anti-synaptophysin, clone SVP-38 (Wiedenmann et al. 1986), anti-substance P, batch 104560 (Keränen et al. 1995), and anti-enkephalin, batch 935 (Soinila et al. 1992). Galanin and VIP immunoreactivities were completely abolished after preincubation with a 1 μM solution of corresponding peptide. According to the documentation provided by the manufacturer, PGP 9.5 antiserum, batch Z00705, recognizes a 24-kD band on Western blotting, representing the PGP 9.5 peptide, in lysates of whole rat brain and human neuroblastoma cell line. After the primary incubation, the glasses were rinsed, incubated with fluorescein-conjugated sheep anti-mouse IgG (1:300; Jackson, West Chester, PA) for GAP-43 antibody and synaptophysin antibody, or with swine anti-rabbit IgG (1:100; DAKO F205, Carpinteria, CA) for PGP 9.5 and NSE antibodies or with rhodamine-conjugated swine anti-rabbit IgG (1:100; DAKO R156) for neuropeptides for 1 hr at RT. After brief rinsing in PBS, the preparation was stained with 0.05% Pontamine Sky Blue for 10 min (Cowen et al. 1985) and mounted in Na-veronal-glycerol mixture. For the double staining experiments, mouse anti-GAP-43 and rabbit-raised neuropeptide antibodies were incubated simultaneously, and the secondary antibodies consecutively. The possibility of the secondary antibodies crossreacting with each other was excluded by incubating the specimen first with one of the rabbit-raised antibodies, followed by anti-mouse secondary antibody. Correspondingly, specimens stained with mouse-raised antibody were incubated with anti-rabbit secondary antibody. No labeling was observed in either case. The specimens were examined with a Leitz Aristoplan fluorescence microscope and photographed. Some specimens were stained according to the avidin-biotin protocol using a Vector ABC kit (Burlingame, CA).

Morphometric Analysis

For each specimen, three consecutive sections from five different regions were randomly selected and stained for GAP-43, PGP 9.5, or NSE. Fields of the circular muscle layer containing transverse sections of nerve fibers were photographed under standardized conditions and the photographs were digitized with a Hewlett Packard Scanjet IIcx scanner. The images were analyzed using the Sigma Scan Pro 4.0 software (SPSS Science; Erkrath, Germany). The area represented by the photograph was 0.33 mm². An intensity histogram of this area was obtained and the threshold of significant intensity was determined by comparing the digitized image with the original photograph. Those pixels whose intensity did not exceed the threshold value were electronically removed. The same threshold value was used for all images. The number of pixels per each profile of transected nerve fiber, the number of profiles per unit area, the average and total intensity, and the perimeter of each profile were measured. For each profile, shape factor SF was calculated using the formula SF = 4π77 X area/(perimeter)². This parameter indicates how circular a profile is, the value for a circle being 1.0 and that of a line approaching 0. To exclude oblique sections, only those profiles were accepted of which the SF was 0.4 or greater. To exclude nonspecific fluorescent aggregates, profiles of which the area was under 4 pixels were omitted. Because some sections contained occasional profiles which, based on their size or shape, were obviously not transected nerve fibers, it was decided to set an upper limit of 300 pixels for the profile area. The mean value for each parameter per each section was calculated and the mean values (± SE) obtained from the five sections of each specimen were averaged. The values of three patients were pooled. Statistical analysis was performed using Wilcoxon's signed rank test and Student's ^t-test. A probability of p < 0.05 was accepted as the significance limit.

Results

Distribution of GAP-43 Immunoreactivity

The distribution of GAP-43 immunoreactivity was similar in the ileum and colon. In the mucosa, GAP-43-immunoreactive nerve fibers were abundant (Figures 1A and 1C). They extended up to the tips of the villi and formed a dense network at the base of the villus. Muscularis mucosae and the submucous layer also contained a network of individual GAP-43-immunoreactive fibers, and the latter layer also showed thick fiber bundles and ganglia containing a few neurons. In the muscle layer, GAP-43-immunoreactive nerve fibers were numerous and in both the circular and longitudinal layer their course was parallel to that of the muscle fibers (Figures 1B and 1D). The muscle cell membrane was nonreactive for GAP-43. In the ganglia, a network of GAP-43-immunoreactive fibers surrounded the neurons, particularly dense in the myenteric ganglia. Therefore, it was difficult to evaluate the intensity of the somatic immunoreactivity. However, we received the impression that the neuronal somata lacked immunoreactivity (myenteric ganglia) or that the immunoreactivity was weak at the most (sub-mucous ganglia) (Figures 3A, 6C, and 6F).

Figure 1

GAP-43 immunoreactivity in the human intestine. (A) Mucous layer of the ileum; (B) circular muscle layer of the ileum; (C) mucous layer of the colon; (D) circular muscle layer of the colon. Bars = 100 μm.

Co-localization of GAP-43 with Neuropeptides

Double labeling studies were carried out to assess the extent of coexistence of GAP-43 and the neuropeptides VIP, substance P, galanin, and enkephalin. GAP-43 partially co-localized with every neuropeptide studied in the nerve fibers. All VIP-, substance P-, and galaninimmunoreactive fibers in all layers also showed GAP-43 reactivity (Figure 2). However, GAP-43-immunoreactive fibers were found that lacked VIP, substance P, or galanin immunoreactivity. Enkephalin-immunoreactive fibers were present only in the muscular layers and they also showed GAP-43 immunoreactivity (not shown). Ganglion cell bodies showing VIP, substance P, galanin, or enkephalin immunoreactivity lacked GAP-43 immunoreactivity (Figure 3).

Figure 2

Double labeling staining of GAP-43 and neuropeptides in the same section in the mucous layer. (A,C,E) GAP-43 immunoreactivity; (B) VIP immunoreactivity; (D) substance P immunoreactivity; (F) galanin immunoreactivity. Arrows indicate fibers where no co-localization exists. Bar = 100 μm.

Figure 3

Double labeling of the submucous ganglia in the colon. (A) GAP-43 immunoreactivity; (B) substance P immunoreactivity. Bar = 100 μm.

Comparison of Synaptophysin, NSE, PGP 9.5, and GAP-43 Immunoreactivities

The lamina propria showed numerous thin synaptophysin- (Figure 4B) and NSE-immunoreactive nerve fibers (Figure 5A). In addition, many cells in the villous core were weakly or moderately stained by both antibodies. Consecutive sections stained with PGP 9.5 antiserum revealed a greater number of reactive nerve fibers (Figure 5B). Another consecutive section stained for GAP-43 showed a dense network of nerve fibers, but no cells were stained (Figure 5C). The number of stained nerve fibers per unit area was even greater than that for PGP 9.5. However, the visual estimation may be biased by the fact the GAP-43-immunoreactive fibers were thicker and more intensely fluorescent than PGP 9.5-immunoreactive fibers, and appeared more sharply demarcated (Figures 5B and 5C). The muscularis mucosae contained fewer synaptophysin-immunoreactive fibers than specimens stained for the other markers. In this layer, no essential differences in the distribution of GAP-43, PGP 9.5, or NSE staining was observed. All four stainings showed immunoreactive nerve fibers around blood vessels. Likewise, all stainings revealed nerve fibers in the peripheral regions of the Peyer's patches. In the muscle layer, all four stainings revealed exclusively nerve fibers, which showed an identical distribution (Figures 4C, 4D, and 5D-5F). Blind visual estimation indicated that the density of GAP-43-immunoreactive fiber profiles was slightly greater than that of the others. As in the lamina propria, the GAP-43-immunoreactive fibers appeared slightly thicker than the PGP 9.5-, NSE-, and synaptophysin-immunoreactive fibers. The submucous and myenteric ganglion neurons showed moderate to strong and even cytoplasmic staining for NSE (Figures 6A and 6D) in addition to the staining in nerve fiber plexuses. The intensity of PGP 9.5 immunoreactivity in the neuronal somata varied from weak to moderate (Figures 6B and 6E), whereas the nerve fibers were intensely reactive (Figure 6E). No synaptophysin immunoreactivity was observed in neuronal somata of the submucous or myenteric ganglia (not shown). In the ganglia, synaptophysin-immunoreactive fibers were less numerous than GAP-43-immunoreactive fibers. No GAP-43 immunoreactivity was observed in the myenteric neurons (Figure 6C) and only weak immunoreactivity was seen in the submucous ganglion neurons (Figure 6F).

For reference, NSE, PGP 9.5, and GAP-43 immunoreactivities were examined in the labial and submandibular salivary glands. Nerve fibers around the acini, tubules, ducts, and blood vessels showed immunoreactivity for all three antigens. The density of NSE-stained fibers was clearly less than that of PGP 9.5- or GAP-43-immunoreactive fibers, which revealed a very rich network of fibers throughout the gland. No essential difference was observed between the latter two stainings.

Morphometric Analysis of GAP-43, PGP 9.5, and NSE Immunoreactivities

On visual examination, GAP-43 immunoreactivity appeared to reveal nerve fibers somewhat better than PGP 9.5 staining, but it was less sensitive in showing neuronal somata than NSE. Therefore, a quantitative comparison between these markers was performed. Synaptophysin was not included in this comparison because it was clearly a less sensitive indicator of nerve fibers and neuronal somata than PGP 9.5 or NSE, respectively. In the colon, the mean number of GAP-43-immunoreactive nerve profiles per unit area was 11% higher than that of PGP 9.5- (p < 0.05) and 33% higher than that of NSE-immunoreactive fibers (p < 0.01) (Figure 7A). The average number of pixels in a single profile was 52% greater for GAP-43 than for PGP 9.5 (p < 0.01) and 60% greater than for NSE (p < 0.01) (Figure 7B). The mean perimeter of the GAP-43-immunoreactive profiles was 24% greater than that of PGP 9.5- (p < 0.01) and 27% greater than that of NSE-immunoreactive fibers (p < 0.01) (not shown). The mean intensity of pixels in each profile was 6% greater in the GAP-43-stained sections than in the PGP 9.5- (p < 0.01) and 8% greater than in the NSE- (p < 0.01) stained sections (not shown). The mean total intensity of the GAP-immunoreactive profiles was 70% greater than that of NSE-immunoreactive profiles (p < 0.01) and 60% greater than that of PGP 9.5-immunoreactive profiles (p < 0.01) (Figure 7C). In the colon, there was no significant difference between the examined parameters of PGP 9.5 and NSE stainings.

In the ileum, the mean number of GAP-43-immunoreactive profiles per unit area was 13% higher than that of PGP 9.5- (p < 0.05) and 121% higher than that of NSE-immunoreactive fibers (p < 0.001) (Figure 7A). The value for PGP 9.5 was 96% higher than that for NSE (p < 0.001). The mean number of pixels in a single profile was 12% higher for GAP-43 staining than for PGP 9.5 (ns) and 84% higher than for NSE (p < 0.001) (Figure 7B). The value of PGP 9.5 was 64% higher than that of NSE (p < 0.001). There was no significant difference in the value for the average intensity of pixels or for the perimeter between GAP-43 and PGP 9.5. For both GAP-43 and PGP 9.5, values were significantly greater than that of NSE (p < 0.01) (not shown). The average total intensity of GAP-43-immunoreactive profiles was 13% higher than that for PGP 9.5 (ns) and 105% higher than that for NSE (p < 0.001). The value for PGP 9.5 was 80% greater than that for NSE (p < 0.001) (Figure 7C).

Figure 4

Immunoreactivity for GAP-43 (A,C) and synaptophysin (B,D) in the human colon. (A,B) Mucosal layer; (C,D) circular muscle layer. Bars = 100 μm.

Discussion

Although GAP-43 is known as a neuronal membrane-bound protein associated with development and regeneration, it is continuously expressed in certain regions of the mature brain (Benowitz et al. 1988,1989), as well as in the autonomic nervous system (Stewart et al. 1992). It has remained unknown whether GAP-43 immunoreactivity correlates with particular neurotransmitters or neuropeptides. The present study demonstrates abundant GAP-43-immuoreactive nerve fibers in the normal human small and large intestine. In previous studies, GAP-43 immunoreactivity has been described in the rat stomach, and small and large intestine (McGuire et al. 1988; Stead et al. 1991; Stewart et al. 1992; Lhoták et al. 1995). GAP-43-immunoreactive nerves have also been reported in the human small intestine and rectum (Sharkey et al. 1990, 1992; Kobayashi et al. 1996). All these studies unanimously report a dense network of GAP-43-immunoreactive nerve fibers in all gut layers. In our study, the distribution of GAP-43-immunoreactive fibers in the normal human colon was similar to that in the small intestine. Our co-localization studies demonstrated that all VIP-, substance P-, galanin-, and enkephalin-containing nerve fibers also showed GAP-43 immunoreactivity. Previously, VIP has been found to co-localize with GAP-43 in the ferret ileum (Sharkey et al. 1990). In the rat jejunal villi, all electron microscopically identifiable nerve profiles were found to be GAP-43-positive (Lhoták et al. 1995). Taken together, these observations suggest that GAP-43 is a universal neuronal marker in the mature enteric nervous system.

Figure 5

Immunoreactivity for three neuronal markers in the human colon. (A,D) NSE; (B,E) PGP 9.5; (C,F) GAP-43. (A-C) Mucosal layer; (D-F) circular muscle layer. Bars = 100 μm.

Figure 6

Immunoreactivity of three neuronal markers in consecutive sections of the enteric ganglia. (A,D) NSE; (B,E) PGP 9.5; (C,F) GAP-43. (A-C) Submucous ganglia; (D-F) myenteric ganglia. Bar = 50 μm.

Submucous ganglion neurons lacked GAP-43 immunoreactivity and myenteric neurons displayed only very weak GAP-43 immunoreactivity. Stewart and co-workers (1992) found no GAP-43 immunoreactivity in the rat enteric ganglia. Sharkey and co-workers (1990) found weak immunoreaction in the ganglion cells in the rat myenteric plexus and also in the neurons in the submucous plexus in the human duodenum. They also demonstrated by in situ hybridization technique that enteric neurons express GAP-43 mRNA. These results indicate that mature enteric neurons synthesize GAP-43, albeit at low levels.

GAP-43 as a neuronal marker has not been compared in the gut with other established neuronal markers, such as PGP 9.5, synaptophysin, or NSE. PGP 9.5 is a cytoplasmic protein localized both in the central and peripheral neurons and nerves (Thompson et al. 1983; Gulbenkian et al. 1987; Lundberg et al. 1988; Wilson et al. 1988). It has been claimed to be a sensitive immunohistochemical neuronal marker, even better than NSE, for the enteric nervous system in whole-mount preparations of mammalian intestine (Krammer et al. 1993,1994). Our visual investigation indicated that GAP-43 reveals a slightly denser nerve fiber network than PGP 9.5 both in the lamina propria and in the muscle layer. The GAP-43-immunoreactive fiber profiles showed greater intensity, were thicker and were more sharply demarcated than those stained with PGP 9.5 antiserum. The better visibility of GAP-43-immunoreactive nerves is likely due not only to greater immunofluorescence intensity per se but also to different intracellular distribution of the two proteins, GAP-43 being membrane-bound whereas PGP 9.5 is soluble. An additional factor contributing to the appearance of GAP-43-immunoreactive fibers is immunoreactivity of satellite cells or nonmyelinating Schwann cells around enteric nerve fibers. This possibility is based on the observation, both in vitro and in vivo, that rat glial cells and nonmyelinating Schwann cells express GAP-43 (Curtis et al. 1992).

Figure 7

Morphometry of neuronal markers in the circular muscle layer of the human intestine (mean ± SEM). The asterisk represents the significance of the difference between GAP-43 and PGP 9.5. The cross indicates the statistical significance between GAP-43 and NSE. The section mark indicates the statistical significance between PGP 9.5 and NSE. ^* p < 0.05, ^** p < 0.01, ⁺ p < 0.05, ⁺⁺ p < 0.01, ^§ p < 0.05, ^§§ p < 0.01.

Comparison of GAP-43 and PGP 9.5 has shown that only a fraction of nerve fibers in normal adult pancreas exhibit GAP-43 immunoreactivity, whereas most if not all nerve fibers are PGP 9.5-immunoreactive (Fink et al. 1994). Likewise, PGP 9.5-immunoreactive nerve fibers in the skin are more numerous than those revealed by GAP-43 (Fantini and Johansson 1992). This is in agreement with the original concept that GAP-43 is associated with developing or regenerating nerves. However, it contrasts with our observation on the human salivary glands in which GAP-43 immunostaining revealed a somewhat denser plexus of nerve fibers than PGP 9.5, as well as with those reported on the rat autonomic nervous system, in which GAP-43 and PGP 9.5 are equally sensitive markers of nerve fibers (Stewart et al. 1992).

NSE is a widely used neuronal marker, although it is also expressed by several neuroendocrine and nonneuronal cells (Marangos et al. 1979). Our observation was similar to that reported previously (Bishop et al. 1982) in showing an NSE-immunoreactive fiber network throughout the lamina propria and muscle layer. Synaptophysin is a neuron-specific protein because it is a structural component of synaptic vesicles. Both NSE and synaptophysin antibodies stained nerve fibers in the human gut, but the profiles were by mere visual examination clearly thinner and less numerous than those stained with GAP-43 or PGP 9.5. The visibility of NSE-immunoreactive nerve fibers was further hindered by nonspecific staining of connective tissue cells in the lamina propria. No nonspecific staining was seen in the muscle layer. Nevertheless, fewer and less intensely stained nerve fibers were revealed with NSE or synaptophysin staining compared with GAP-43 or PGP 9.5 staining.

In contrast to nerve fiber staining, NSE was a superior marker of ganglion neurons. All ganglion neurons showed strong cytoplasmic immunoreactivity. Somatic PGP 9.5 immunoreactivity varied in intensity, and GAP-43 revealed only vaguely reactive neuronal somata at the most. The poor staining by the GAP-43 antibody is probably due to rapid transportation of the protein into the axons to be incorporated into the axolemma.

Quantitative analysis of the nerve fibers in the circular muscle layer verified our visual observations for the order of sensitivity of the three markers: GAP-43>PGP 9.5>NSE. In both the colon and ileum, GAP-43 revealed over 10% more nerve fibers per unit area than the other two markers, indicating that GAP-43 staining reveals fibers unstained by PGP 9.5 antiserum. The difference in average number of pixels per single profile, a parameter correlating with profile size, was significant in the colon, and the trend was similar in the ileum. A similar result was obtained when average profile perimeters were compared. These results suggest that the fibers revealed by GAP-43 staining are thicker than those stained by the PGP 9.5 or NSE antibody. Total intensity, as defined in the present study, sums the intensity values of each pixel in a profile and is therefore sensitive to changes in profile size and/or intensity. Therefore, the difference in total intensity of a single profile between GAP-43-immunoreactive and the other two markers is expected and can be explained by the differences in the number of pixels per profile. In fact, the values for average intensity of profiles were essentially similar for each antibody, the difference between GAP-43 and the other two being only about one tenth of the difference in total intensity values. In summary, our results suggest that GAP-43 immunoreactivity reveals more numerous and thicker nerve fibers than PGP 9.5 or NSE immunoreactivity, although the true staining intensity is essentially similar for the three markers. In contrast, NSE is a superior marker for neuronal somata.