Basal forebrain cholinergic cell attachment and neurite outgrowth on organotypic slice cultures of hippocampal formation.

4Distributions of somata and neurites of cholinergic neurons were studied after seeding dissociated cells onto organotypic slice cultures. Slice cultures were made from hippocampal formation and adjacent cortical regions from rats or mice. Dissociated cell suspensions of basal forebrain tissue from rat or mouse fetuses were seeded onto the slice cultures. Combined cultures were maintained for 1^21 days in vitro . Cultures processed for acetylcholinesterase (AChE) histochemistry demonstrated non-random patterns of cholinergic cells and their neurites. Labeled cells appeared most frequently in the molecular layer of the dentate gyrus, and in the deeper layers of cortical regions adjacent to the hippocampus. Neurites extending from these labeled cells appeared to target the dentate molecular layer and the cortical subplate layer. By 4 days in vitro , AChE-positive basal forebrain cells display several short and thick neurites that appear to be dendrites, and one long process that appears to be an axon. By 5 days in vitro , dendrites are well developed; by 7 days the presumed axon has extended widely over the cortical target zone. These neurites are maintained through 3 weeks in culture. Distributions of cells varied with the age of the slice. AChE-labeled cells were not seen overlying hippocampal tissue when dissociated cells were seeded on slice cultures made from day 0 rats, but a few labeled cells were seen when seeded on slices from day 2 rats. Clear non-random patterns of labeled cells and neurite outgrowth were seen on slice cultures from day 5 or older pups. The non-random distribution seen with AChE-positive neurons was not seen using other techniques that labeled all cells (non-selective £uorescent labels) or all neurons; these techniques resulted in labeled cells scattered apparently homogenously across the slice culture. These studies demonstrate a non-random pattern of attachment or di¡erentiation of basal forebrain cholinergic neurons when these cells are seeded onto cultured cortical slices; this pattern mimics the normal patterns of basal forebrain cholinergic projections to these cortical regions. These data suggest that the factors that normally guide basal forebrain-derived cholinergic axons to their target cells in vivo are present and detectable in this model system. D IBRO.

It is well recognized that neuronal circuitry is established through the formation of appropriate synaptic connections in the developing brain. Recent research has demonstrated that formation of these connections is likely to be in£uenced by a variety of factors, including attractive, repulsive, and guidance cues (Goodman and Shatz, 1993;Johansen et al., 1994;Lander, 1990;Landmesser, 1994). These factors serve to bring the axons and dendrites of two di¡erent cells into close proximity so that cell^cell interactions can allow the growing axons to recognize their target neurons (Goodman and Shatz, 1993;Johansen et al., 1994;Lander, 1990;Landmesser, 1994).
Studying these putative cell^cell recognition processes in the whole developing animal is a challenging problem that has prompted the use of more reduced in vitro approaches (Alderson et al., 1990;Amaral and Witter, 1995;Baratta et al., 1996a,b;Robertson, 1992, 1993;Dreyfus, 1989;Ha et al., 1996;Hartikka and Hefti, 1988). Recently, we have introduced a method in which dissociated cells are seeded onto organotypic slice cultures of known target tissue . Some of the seeded neurons attach and di¡erentiate on the slice culture. The locations of the seeded cells, and the patterns of their axonal growth, can be used as indicators of the distributions of factors within the slice tissue that promote, guide, or repulse cell attachment, di¡erentiation, or outgrowth .
The goal of the present investigation was to determine whether dissociated BFCN somata and their neurites are distributed in non-random patterns that mimic the normal patterns of cholinergic ingrowth seen in vivo. Such patterns would be interpreted as indicating the presence of local factors that attract, repel, or nurture cholinergic elements. Portions of these data have been presented previously in abstract form (Haraldson et al., 1997).

Animals
Fetal and infant rats and mice were used in these experiments. Rats were of the Sprague^Dawley strain from Simonsen Labs, and included fetuses of gestational age 15^17 days (G15^17) and newborn pups of postnatal age 4^7 days (P4^7). Both wild-type (C57BL/6) and transgenic (see below) newborn (P5^7) and fetal (G15^17) mice were used. All experiments were carried out to minimize animal su¡ering and to reduce the number of animals used. All use of animals was in accordance with the National Institute of Health Guide for the Care and Use of Laboratory Animals and was approved by the University of California, Irvine, CA, USA, Institutional Animal Care and Use Committee.

Materials
All reagents were purchased from Sigma Chemical (St. Louis, MO, USA) unless speci¢ed otherwise.

Organotypic slice culture preparations
Organotypic slice cultures using the 'membrane technique' (Ga «hwiler, 1981;Ga «hwiler et al., 1998;Stoppini et al., 1991;Yamamoto et al., 1989Yamamoto et al., , 1992 were prepared from brains of P0^7 rat or mouse pups, using sterile techniques under a laminar £ow hood. Animals were cryoanesthetized and killed by decapitation, and the brains were quickly removed. Transverse or horizontal 400-Wm slices of brain tissue were cut on a Vibratome and placed into cold Dulbecco's modi¢ed Eagle's medium (DMEM). Under a dissecting microscope, a ¢ne knife was used to isolate the hippocampus in transverse sections, or the hippocampus and adjacent entorhinal cortex and perirhinal cortex in horizontal sections. Each slice was placed alone on a 30mm Millicell culture insert (Millipore, MA, USA). Four culture inserts with tissue slices were placed into a 100-mmU20-mm Petri dish containing 7 ml of culture medium. The medium was made with 100 ml basal Eagle medium (with Earle's salts), 50 ml Earle's balanced salt solution, 50 ml heat-inactivated horse serum, 1 ml 200 mM L-glutamine, and 2 ml 50% glucose.
Slice cultures were maintained in a 36 ‡C incubator with 5% CO 2 for 1^4 days before the addition of suspensions of dissociated cells.

Dissociated cell culture preparations
Pregnant rats or mice of gestational days 15^17 were deeply anesthetized with sodium pentobarbital (50 mg/kg). Anesthetized fetuses were surgically removed from the uterus and decapitated. The brains were then removed and placed in cold DMEM. Under a dissection microscope the basal forebrain region, including septum and diagonal band, was exposed and removed surgically, using sharp forceps (Ha et al., 1996;Hartikka and Hefti, 1988). Basal forebrain tissue pieces were incubated in 0.09% trypsin in DMEM at 36 ‡C for 30 min. The tissue was centrifuged at 3000 r.p.m. for 5 min, and then the supernatant was removed and replaced with serum-based medium. The tissue was dissociated by trituration with ¢re-polished ¢ne barrel pipettes and suspended in the culture medium. The cell suspension had a concentration of approximately 2U10 2 cells/Wl. A volume of 5^20 Wl of the cell suspension was carefully pipetted onto the previously prepared slice cultures. These combined cultures were allowed to survive for periods of 1 day to 3 weeks.

Prelabeling of dissociated cells
In some cases cell suspensions were prelabeled so that cells could be detected after seeding on slice cultures. After incubation with trypsin, cell suspensions were mixed for 20 min at 37 ‡C with either rhodamine-labeled £uorescent microspheres (Fluorospheres; Molecular Probes, Eugene, OR, USA) at 1 Wl/ ml, or Vibrant DiI cell tracker (Molecular Probes, Eugene, OR, USA) at 5 Wl/ml. After labeling, cells were centrifuged and washed ¢ve times at 5 min each before being suspended again in tissue culture medium and seeding onto organotypic slices.
In other cases, basal forebrain tissue from transgenic mice engineered to carry their own endogenous marker was dissociated; these included strains that strongly express either green £uorescent protein (GFP; Okabe et al., 1997) or L-galactosidase (L-gal). As described by Davies et al. (1999) the GFP expressing transgenic mice were maintained on a C57BL/6N background, with the gene for the GFP protein under control of a chicken L-actin promoter (Okabe et al., 1997). In these animals, virtually all neurons, along with other cells, express GFP. Other transgenic mice were the genetically engineered tattler IV (an acronym for T-alpha 1-tubulin-tau-LacZ-expresser) strain (Murray et al., 2002), in which neurons express a tau^L-gal fusion protein under direction of the neuron-speci¢c T-alpha-1 alpha tubulin promoter (Gloster et al., 1994). These mice were maintained on an outbred CD-1 background.
After ¢xation with 4% paraformaldehyde, the DiI and £uorescent latex microsphere (FLM)-labeled cells were detected directly using the £uorescence microscope, using a rhodamine ¢lter. Similarly, GFP-positive cells were detected directly, using a £uorescein ¢lter. The L-gal-positive cells from tattler mice were detected immunocytochemically. After ¢xing with 4% paraformaldehyde in 100 mM sodium phosphate-bu¡ered saline (PBS) for 1 h, cultures were rinsed in PBS and processed for immunocytochemistry using a primary polyclonal antibody to L-gal (5 prime 3 prime) at 1:1000 followed by either a biotinylated or a rhodamine-labeled secondary antibody at 1:200. The biotinylated secondary antibody was followed by the ABC Elite kit (Vector) for 1 h each and then the reaction was visualized by 0.02% 3,3P-diaminobenzidine with 0.001% H 2 O 2 in Tris^saline for 6^10 min.

AChE histochemistry
Cultures were ¢xed by immersion in 4% paraformaldehyde and 0.5% glutaraldehyde in 100 mM sodium phosphate bu¡er overnight at 4 ‡C and processed for AChE histochemistry as described by Tago et al. (1986). Cultures were rinsed in three changes of 100 mM maleate bu¡er (pH 6.0) for 10 min each, and then incubated in 50 ml maleate bu¡er, with 30 Wl 86.5 mM acetyl-thio-choline iodide, 100 Wl 30 mM tetra-isopropyl pyrophosphoramide and 500 Wl Karnovsky and Roots solution (made up from 2.5 ml 100 mM sodium citrate, 5 ml 30 mM copper sulfate, 8 mg potassium ferricyanide and 42.5 ml 100 mM maleate bu¡er). After 5 h incubation in the dark, the cultures were rinsed in three changes of 50 mM Tris bu¡er (pH 7.5) and then reacted in 0.05% 3,3P-diaminobenzidine, 0.003% hydrogen peroxide and 0.2% nickel ammonium sulfate in 50 mM Tris bu¡er for 5^10 min when stained ¢bers appeared. Cultures were rinsed in several changes of 50 mM Tris bu¡er, mounted on gelatin-coated slides, dehydrated in ethanol, and cleared in xylene before coverslipping.

Analysis
Histochemically stained cultures were examined under bright-¢eld microscopy and documented photographically using Kodak technical pan ¢lm. Images from negatives were scanned and imported into Adobe Photoshop. Individual images were adjusted for consistent brightness and contrast for each photographic plate; no other modi¢cation of the images was done.
In addition to the photographic documentation, drawings were made of the location of labeled AChE-positive neurons and axons on composite drawings of the cultures. Drawings were made of individual AChE-stained neurons, using a camera lucida and either a 40U or 63U objective lens.

Distribution of AChE-positive neuron somata on slice cultures
The AChE-positive neurons were distributed in a nonrandom pattern across the cultured slices of hippocampus and adjacent regions. Figure 3A presents a composite camera lucida drawing showing the distribution of all these neurons from a total of 20 cultures made from horizontal sections through the temporal pole, while Fig.  3B presents a similar composite from a total of 15 cultures made from transverse sections of the hippocampus. In the horizontal slices (Fig. 3A) labeled cells were distributed mainly on the dentate gyrus of the hippocampus and on regions along the inner layers of perirhinal and entorhinal cortices. Counts of labeled cells on cultures made from horizontal sections and from transverse sections reveal that AChE-positive cells were found on the dentate gyrus at approximately two to three times the density (labeled cells per unit area) as cells found over the hippocampal CA ¢elds. Within the dentate gyrus, the AChE-positive cells tend to appear over the inner molecular layer. Furthermore, counts revealed that labeled cells were found over the deep layers of perirhinal cortex at approximately twice the density as was seen over super¢cial layers of neocortex.

Distribution of neurites
Neurite outgrowth also appeared in non-random patterns. The photomicrographs in Fig. 4 illustrate the ¢nding that AChE-positive neurites appear concentrated in the dentate gyrus in general (Fig. 4A), and in particular in the dentate molecular layer (Fig. 4B^D). Staining of neurites in the hilus varied considerably between cultures, but AChE-positive neurites were a consistent feature in the molecular layer. Figure 3 also presents composite drawings of AChEpositive BFCN neurite outgrowth for cultures made from horizontal (Fig. 3C) and transverse (Fig. 3D) slices through the hippocampal formation and adjacent regions. The pattern of AChE-positive neurites is similar to the pattern of AChE-positive somata. Examples of distributions of AChE-positive neurite outgrowth also are illustrated in the camera lucida drawings presented in Fig. 5.

Development and maintenance of neurites
The development and di¡erentiation of BFCNs and the growth of their neurites were studied in cultures maintained for 1 day to 3 weeks. Examination of AChEstained cultures 1 or 2 days following the seeding of dissociated basal forebrain cells revealed no detectable BFCNs on the tissue (data not illustrated). Three days following the seeding of dissociated cells on the cultured slices, a few small and lightly AChE-positive cells with poorly developed neurites could be seen (Fig. 2C). By 4 days in vitro (DIV) however, AChE-positive neurons were more prominent (Fig. 2D), with clearly visible somata and AChE-positive neurites. Most of the neurons at this stage had relatively short neurites, which rarely extended farther than 50 Wm from the soma. In cultures of more than 4 DIV, labeled cells had several neurites, including a single thin and long (greater than 100 Wm) axon-like process and several shorter and thicker dendrite-like processes. By 7 DIV, AChE-positive neurons were commonly observed; these cells displayed larger somata and more extensive neurite outgrowth (Fig. 1E) than those from shorter culture periods. Continued growth of AChE-positive neurites was apparent in cultures maintained for longer than 1 week in culture ( Fig. 2A, B) with an accompanying increase in the intensity of the AChE histochemical reaction product. Cultures maintained for 3 weeks displayed strongly AChE-positive neurons (Fig. 1F), with neurites that were much longer and more complex (Fig. 1F). Although the morphological features of individual neurons appeared more complex in older cultures, the distributions of AChE-positive cells across the tissue slice were similar to those observed with shorter culture periods. No evidence of degeneration of AChE-positive neurons was detected with the longer (3 weeks) culture periods.
With culture periods of 4 days up to 3 weeks in vitro, patterns of neurite outgrowth appeared similar to those seen with shorter culture periods, although the AChEpositive neurites appeared more profuse at 3 weeks than at 7 days. Axon projections were maintained after 3 weeks in culture, indicating that axonal innervation into dentate gyrus, entorhinal cortex, and inner cortex was not transient.

Development of adhering or growth promoting aspects of the hippocampus
The non-random distribution of BFCNs on the cultured slice suggests that the seeded BFCNs are attaching, surviving, or di¡erentiating by some factor(s) produced by the slice tissue. A set of studies examined the role of maturation of the hippocampal tissue on the patterns of BFCNs. Figure 6 presents examples of slices of hippocampus taken from animals at P0 (Fig. 6A), P2 (Fig. 6B) and P7 (Fig. 6C). BFCNs from the same age fetal tissue (gestational day 17) were seeded onto each of the slices after the slice had been maintained 1 day in culture; the combined cultures then were maintained for an additional 7 days. Note that no AChE-positive BFCNs were seen on the slice taken at P0 (although a few labeled cells could be detected around the edge of the cultured slice). A few AChE-positive cells were seen on the slice taken at P2, and the stereotypical pattern of BFCNs was seen on the slice taken at P7. These data suggest that some aspect of the hippocampal formation must mature prior to the addition of the dissociated basal forebrain cells, in order for the BFCNs to attach, survive, or di¡erentiate on the cultured slice.
Does the non-random distribution of AChE-positive neurons result from initial non-random attachment of neurons?
The data presented in the previous section demon-strate that postnatal maturation of hippocampal tissue is important for the non-random distribution of AChE neurons. Selective adhesive factors demonstrated in other models (Emerling and Lander, 1994;Fo « rster et al., 1998;Fo « rster, 2001) may be important here. Dissociated cells prelabeled with the DiI cell tracker or with FLM were seeded onto slices from P5 animals, left undisturbed for 2 h, and then gently rinsed with culture medium. These cases showed no evidence of the presence of labeled cells 2 h after washing, and no evidence of AChE-labeled cells several days after washing (data not illustrated). Thus, these data suggest that initial cell attachment does not play an essential role in the later distribution of AChEpositive cells.
Is the non-random distribution displayed by AChE-positive neurons simply one example of the non-random attachment of all cells?
The data indicate that AChE-positive cholinergic neurons are distributed non-randomly on slices of hippocampal formation and neocortex. Experiments with prelabeled dissociated basal forebrain cells explored whether this pattern was characteristic of cholinergic neurons speci¢cally, or whether all cells would display this pattern. The photomicrographs in Fig. 7 present examples of labeling of cells with a variety of techniques, including cells from transgenic animals that carried endogenous labels (L-gal: Fig. 7B or GFP: Fig. 7E) and cells that were labeled by exposure to direct cell labeling techniques while in dissociated cell suspension (DiI: Fig. 7D or FLM: Fig. 7F). Cells labeled by each of these methods did not display the pattern typical of the AChE-positive neurons, but rather were distributed apparently randomly across the culture. Related studies using double labeling techniques demonstrated that the majority of FLM-positive cells were likely to be microglia, as they displayed a positive reaction to the tomato lectin (Eliason et al., 2002). Vibrant DiI-labeled cells appeared to be both neuronal and glial, with glial cells (co-labeled either by tomato lectin or GFAP immunocytochemistry) being the majority.
Both neuronal (Fig. 7E) and non-neuronal cells from the GFP expressing transgenic mice displayed GFP £uorescence. GFP-positive cells with neuronal morphological features appeared distributed homogenously across the slice culture, although these labeled cells appeared slightly more frequently over the region of the hilus of the dentate gyrus, and slightly less frequently over the dentate gyrus molecular layer. The L-gal immunoreactive cells (Fig. 7B) virtually all had morphological features of neurons, and displayed a homogenous distribution pattern similar to that of the GFP-labeled cells.

DISCUSSION
The present results demonstrate that AChE-positive BFCNs and their neurites are found in non-random distributions after dissociated basal forebrain cells are seeded on slice cultures of hippocampus and adjacent tissue. This ¢nding is important because the distributions of BFCN somata and the neurites in this experimental test situation may provide information regarding mechanisms by which cholinergic axons normally ¢nd their targets in the cerebral cortex.

Technical considerations
Histochemical techniques for detecting AChE have long been used to identify BFCNs (Butcher, 1983;Distler and Robertson, 1992;Lewis and Shute, 1967;Robertson et al., 1997;Wenk et al., 1980;Woolf, 1991). While a number of instances have been documented in which AChE is expressed by non-cholinergic cells (Green¢eld, 1984;Layer and Willbold, 1995;Lehman and Fibiger, 1979;Levey et al., 1984;Robertson et al., 1988;Robertson and Yu, 1993;Silver, 1974;Small, 1990), co-localization Fig. 6. AChE-stained slice cultures made from transverse slices of brains taken at P0 (A), P2 (B), and at P7 (C). Note that while AChE-positive BFCNs (arrow heads) are found at the edge of the P0 culture, they appear absent from the P0 hippocampal tissue. A few AChE-positive cells are seen over the hippocampus and dentate at P2 (B), and are more prominent on the culture taken at P7. Calibration bar = 500 Wm.
E. S. Tsai et al. 822 studies have demonstrated that virtually all BFCNs express AChE Woolf, 1991). AChE is expressed by the neuronal somata and dendrites, and also by the axons of cholinergic neurons, making it a reliable and valid marker for revealing the presence and morphological features of BFCNs. AChE-positive neurons also are a normal component of cortical regions, including the hippocampus, entorhinal cortex and neocortex (Levey et al., 1984;Lysakowski et al., 1989;Robertson et al., 1988Robertson et al., , 1990Seress et al., 1987). Interestingly, these AChE-positive neurons appear not to be cholinergic (Levey et al., 1984). However, morphological criteria can be used to distinguish these AChE-positive intrinsic cortical neurons from the seeded basal forebrain-derived AChE-positive neurons; the cortical neu-rons display smaller somata, more limited dendritic ¢elds, and much lower levels of AChE histochemical activity than do the seeded BFCNs. With the techniques employed in the current study, few of the intrinsic AChE-positive cells were stained in most of the cultures. The greater level of AChE histochemical staining of the BFCNs may be due to their larger size and higher levels of endogenous AChE, and because the BFCNs were seeded on top of the slice and consequently had more direct access to the histochemical reagents.
The seeding of dissociated cells onto slices of potential target tissue is a powerful method of investigating cellĉ ell interactions in developing brain (Emerling and Lander, 1994;Fo « rster et al., 1998;Fo « rster, 2001;Halloran and Kalil, 1996;Laywell et al., 1996; Robertson et al., 1997). In normal brain development, of course the somata and dendrites of basal forebrain neurons do not come in contact with cortical structures. It is of particular interest, then, that BFCNs not only display neurite (presumed axonal) outgrowth that forms patterns similar to the patterns formed by basal forebrain-derived cholinergic a¡erents, but also that the BFCN somata themselves show patterns similar to those of the normal cholinergic axonal projections. Thus, the patterns displayed by BFCNs after seeding onto cultured slices are remarkably similar to the pattern displayed by BFCN axonal ingrowth in vivo, suggesting that the mechanisms that serve to attract, retain, or nurture the cholinergic axons to their target sites can be in£uential on the whole neuron.

Normal pattern of basal forebrain cholinergic a¡erents to cerebral cortex
Cholinergic neurons in the basal forebrain, including the medial septum, diagonal band complex, substantia innominata, and medial globus pallidus, send axonal projections to the cerebral cortex in a loosely organized topographic pattern (McKinney et al., 1983;Wenk et al., 1980;Woolf, 1991). The cells of the medial septum project to the hippocampal formation, the diagonal band to the medial cortex and olfactory bulb, and the substantial innominata and medial globus pallidus project to the lateral neocortex (Amaral and Kurz, 1985;Calarco and Robertson, 1995;Luiten et al., 1987;McKinney et al., 1983;Mesulam et al., 1983;Wenk et al., 1980;Woolf, 1991). Although the basal forebrain cholinergic projections to cortex have often been viewed as poorly organized or 'non-speci¢c', regional di¡erences in laminar patterns of innervation between cortical regions have been demonstrated (Lysakowski et al., 1989;Mechawar et al., 2000;Mechawar and Descarries, 2001).
The projection from medial septum to the dentate gyrus displays one of the more clearly layered patterns of termination by cholinergic a¡erents (Amaral and Witter, 1995;Baratta et al., 1996a,b;Deller, 1998;Deller et al., 1999;Fo « rster et al., 1997;Makuch et al., 2001;Storm-Mathisen and Blackstadt, 1964;Super and Soriano, 1994;Zimmer et al., 1986), and it is this layered quality that makes this region an ideal model system for this study. Septal cholinergic projections to the dentate gyrus terminate principally in the inner molecular layer, although termination also occurs in the outer molecular layer and, to some extent, within the hilus (Amaral and Witter, 1995;Deller et al., 1999;Makuch et al., 2001;Milner et al., 1983;Super and Soriano, 1994). Patterns of BFCN distribution, as well as the pattern of AChEpositive neurite outgrowth, mimic the normal pattern of cholinergic a¡erents. It should be noted that septal projections to the dentate gyrus are not exclusively cholinergic (Amaral and Kurz, 1985;Manns et al., 2001;. GABA-ergic basal forebrain neurons also project to the hippocampal formation (Ko « hler et al., 1984;Peterson et al., 1987) and appear to target specifically the GABA-ergic local circuit neurons in the hilus (Gulya ¤s et al., 1999). At present, we have no information regarding the distributions of basal forebrain GABAergic or glutamatergic neurons following the seeding procedure.

Mechanisms leading to the non-random distribution
The non-random distribution of BFCNs after several days in culture may stem from di¡erential attachment of the BFCNs, di¡erential survival, or region-speci¢c di¡erentiation of the AChE-positive phenotype. This issue of the basis of the di¡erential distribution of AChE-positive BFCNs is important for understanding mechanisms that guide the formation of connections in the developing or regenerating nervous system, but at the present time it is di⁄cult to distinguish between these possible mechanisms; indeed, more than one of these may be operating.
As a ¢rst possibility, cell attachment or adhesion is likely to be important. Previous work by Emerling and Lander (1994) on distributions of carbocyanine dyelabeled thalamic neurons seeded onto slices of neocortex, demonstrated that the thalamic neurons were found in non-random patterns in short term (a few hours) cultures, suggesting that selective adhesion factors may lead to di¡erential attachment. However, the present experiments reveal little evidence for the initial role of attachment factors for basal forebrain neurons, as all neurons could be e¡ectively removed from the slice with a very low turbulent wash.
Possibly countering the action of putative attraction factors, we have shown that repulsive factors may also be important in some cases. Repulsion factors were suggested by evidence showing populations of neurons that normally do not project to the cultured slice tissue do not adhere or survive when seeded in this protocol . For example, basal forebrainderived cholinergic neurons are not detected when seeded onto slices of striatum, although tyrosine hydroxylase expressing neurons from ventral midbrain are detected readily. Thus, a combination of factors that attract some cells and repel other cells may play important roles in the non-random distribution of the seeded dissociated cells. Recent work has demonstrated that the semaphorins (Che ¤dotal et al., 1998;Steup et al., 1999) act as repulsive agents for some hippocampal a¡erents, and it may be that the absence or reduced numbers of BFCNs on some regions of cortex is related to the expression of a repulsion factor.
Adhesion factors have been clearly demonstrated for other systems (Emerling and Lander, 1994;Fo « rster et al., 1998;Fo « rster, 2001). Although the present data indicate that adhesion factors of the type demonstrated by Fo « rster and colleagues (Fo « rster et al., 1998;Fo « rster, 2001) may not play a role in initial attachment, this does not mean that adhesion factors do not develop over longer periods of time. Cell adhesion molecules have received considerable attention in recent years (Lander, 1990;Landmesser, 1994;Pinksta¡ et al., 1999;Wilson and Snow, 2000). Although candidate molecules, including the integrins, semaphorins, and chondroitin sulfate proteoglycans, are present in the hippocampus, little information is currently available regarding whether these molecules may be functional in this case. It may be of interest that the AChE that is expressed by the BFCN axons has itself been implicated as serving a cell adhesive function (Layer and Willbold, 1995;Robertson et al., 1998;Robertson and Yu, 1993;Sharma et al., 2001;Small, 1990) although recent work by Mesulam and colleagues (Mesulam et al., 2002) demonstrates that septal projections to dentate gyrus develop normally in AChE knock-out mice.
The AChE-positive phenotype does not become detectable until after 2^3 DIV in these cultures. Because the dissociated cells were taken from brains of animals at gestational day 17, this 2^3-day delay may simply re£ect the normal sequence of development of these neurons (Amaral and Witter, 1995;Brady et al., 1989;Ha et al., 1996;Milner et al., 1983;Shingai et al., 1990;Thal et al., 1991). However, it remains possible that the development of the cholinergic phenotype occurs preferentially in cells that have settled down in a supportive environment or have established appropriate adhesive contacts on the slice. Other studies have reported expression of the cholinergic phenotype in dissociated cultures (Dreyfus, 1989;Hartikka and Hefti, 1988;Li et al., 1995;Nonomura et al., 1996;Shingai et al., 1990), however, the presence of exogenously applied neurotrophic factors or the presence of normal target cells is important for expression of the cholinergic phenotype and for survival (Dreyfus, 1989;Ha et al., 1996;Hartikka and Hefti, 1988;Li et al., 1995;Morse et al., 1993;Nonomura et al., 1996;Shingai et al., 1990). Grown in isolation, without target contacts or supplemental neurotrophins, the BFCNs do not thrive.
Di¡erential survival or di¡erential expression of phenotype may also be important here. Two of the neurotrophins, neurotrophic factor 3 (NT-3) and brain-derived neurotrophic factor (BDNF), have been shown to be expressed by dentate gyrus granule cells (Lauterborn et al., 1994;Makuch et al., 2001). Indeed, both the spatial and the temporal patterns of expression of NT-3 and BDNF mRNAs correspond closely to the spatial and temporal patterns of septally derived cholinergic a¡erents to the dentate gyrus, and to the patterns of BFCN somata and neurites after seeding. If BFCN axons are targeting the cells that express, and presumably release, NT-3 or BDNF, then those cells that contact the dentate granule cells may be stimulated by one or both of these neurotrophins to di¡erentiate to the AChE-positive phenotype. If this is the case, then BFCNs seeded onto slices of hippocampus taken from animals genetically engineered to eliminate the gene for NT-3 or BDNF would be expected not to form the pattern characteristic of the wild-type animal.

Conclusion
The distributions of BFCNs, and their neurites, following seeding onto cultured slices of hippocampus and adjacent tissues are remarkably similar to the distributions of basal forebrain-derived cholinergic axons in the normal brain. It may be that this characteristic pattern results from one or more factors, including selective attachment and selective di¡erentiation stemming from response to neurotrophin production by cells in the cultured slice. Seeding of dissociated cells onto cultured slice tissue may provide a useful means of investigating mechanisms that normally determine the patterns of termination of a¡erent projection systems.