Cytology and Circuitry

Stanley J. Watson, Jr. and William E. Cullinan

This chapter is a discussion of the principles involved in the application of anatomical–biochemical methods to the study of the brain. It emphasizes the cell biological aspects of neuronal function, as well as neuronal operation in the context of multicellular circuits. The aim of this chapter, in conjunction with the preceding chapter on molecular biology and the following on neurochemistry, is that of (a) an appreciation of the larger principles governing genomic functioning at the cellular level and (b) a sense of the essentials of neurotransmission within neuronal circuits (see also Electrophysiology).

The initial section is a brief review of molecular aspects of neurochemistry, enabling the reader to focus on key biochemical elements used to study cellular regulation. For example, it is important to appreciate the functions of different types of neuronal proteins in order to readily understand the techniques used to study them in cell and circuit contexts. Among such sets of proteins are the biosynthetic enzymes (e.g., tyrosine hydroxylase), receptors (e.g., D2 dopamine receptor), and peptide precursors (e.g., pro-opiomelanocortin). Equally important is an understanding of the related actions of the gene encoding them, their transcription into messenger RNA (mRNA), and translation into protein. The body of the chapter focuses on the conceptual framework underlying the methods reviewed, rather than being a technical "cookbook." We will attempt to present a clear view of the rationale behind the method, followed by a brief analysis of particular advantages and limitations. Four main approaches will be emphasized: receptor autoradiography, immunocytochemistry, in situ hybridization histochemistry, and neuroanatomical tract-tracing. We hope to convey how the complementary nature of information gained through the application of these methods has helped move neuroscience research to a level of power and sophistication previously unattainable. The goal of this chapter is then really twofold: to briefly highlight key elements of neuronal cell biology, and to indicate the breadth of available tools for studying neuronal regulation in an anatomical context.

Among the unique components of neurons are those focused on generating an action potential, producing neurotransmitters, and receiving transsynaptic signals. Most of neurobiology and much of biological psychiatry thus depends in part on the analysis of these elements of brain functioning. At its core, the neuron is specialized in its ability to be "excited" by incoming signals and to pass that (modified) information on to other cells. As we look at a typical neuron (), some of the essential elements for this communication can be seen, including a host of specialized proteins. If we begin with a vesicle in a typical dopamine neuron in the substantia nigra, for example, we see that it contains a set of synthetic enzymes—in this case tyrosine hydroxylase and aromatic amino acid decarboxylase—which are responsible for the conversion of the amino acid tyrosine into dopamine. Furthermore, this dopamine vesicle might contain a peptide precursor (e.g., pro-neurotensin) and the enzymes necessary for its processing (as many as 3–6 different types of peptidases) into the final active peptide product, neurotensin. As this vesicle is moved from the cell body to the terminal by an elaborate transport system, these enzymes act to produce their final transmitter products. At the synapse, vesicles are accumulated so that at the appropriate electrochemical signal (involving another major class of proteins) their products are released into the synaptic cleft. The postsynaptic cell "senses" the released dopamine and neurotensin by their action on specific dopamine and neurotensin receptors, which had been previously synthesized and inserted into the membrane. It is clear that many other types of proteins are active in this signaling process. For example, transduction from the receptor to a series of second- and third-order messenger proteins involves a host of molecules, among them channels, G proteins, cyclases, kinases, and transacting factors (see Cholinergic Transduction, Signal Transduction Pathways for Catecholamine Receptors, and Serotonin Receptors: Signal Transduction Pathways). Beyond this array of postsynaptic activity are a series of residual actions on the presynaptic side. For example, monoamines are usually removed from the synapse into the cytosol by a transmitter-specific transporter (see Cholinergic Transduction and Norepinephrine and Serotonin Transporters: Molecular Targets of Antidepressant Drugs); there the transmitter (dopamine) can be moved into the vesicle for re-use (an event accomplished by yet another specific protein), or it can be metabolized by degradative enzymes (e.g., monoamine oxidase). Dopamine remaining in the synapse can also activate another dopamine receptor subtype on the dopamine-producing cell. This dopamine "autoreceptor" inhibits further dopamine release into the synapse (see Molecular Biology, Pharmacology, and Brain Distribution of Subtypes of the Muscarinic Receptor).

The foregoing discussion was presented in order to highlight the diversity of proteins needed for effective neurotransmission. It is indeed these very proteins which can be used to study neuronal regulation in brain. The reader may notice that this discussion began with a complement of mature proteins in neurons. Neurobiology texts often begin there in part because this was basically the extent of our knowledge regarding them 10–15 years ago. As the fields of cell biology and molecular biology have grown, it has become clear that all cells use the same fundamental genetic and protein synthetic machinery. The proteins critical to neuronal functioning are thus the products of standard transcriptional and translational processes. We now turn to a brief description of some of these early events, before describing the aforementioned histochemical and anatomical methods.

As seen in , the double-stranded DNA of the gene is transcribed into a long, single-stranded RNA molecule. This primary RNA transcript [itself a member of a family of RNAs known as heteronuclear RNA (hnRNA)] is a full copy of the gene containing introns and exons alike. Introns are parts of the gene which are not represented in the final mature mRNA; exons are the parts of the gene which comprise the mRNA. After production of this primary transcript, it is then acted upon by a series of splicing enzymes which remove all introns, so all that remains is a series of connected exons. Prior to moving out of the nucleus, this RNA is capped at its 5¢ end with a special nucleotide, and a 3¢ tail of adenines is attached (often 100–200 A's in length). It is now a mature mRNA and passes into the cytoplasm for translation into protein. Free mRNA forms a complex with ribosomal and transfer RNA machinery, ultimately leading to the formation of the peptide chain. Thus, regardless of whether the focus of study is a protein in the form of an enzyme, peptide precursor, or receptor, it is possible to examine the process of its synthesis in the cell body by quantitating hnRNA (to estimate rates of gene transcription) or, much more commonly, mRNA (to infer the ability of the cell to rapidly synthesize protein).

The study of a receptor protein could be carried out in a number of ways. The protein could be localized by the use of antibodies (immunocytochemistry), its mRNA could be detected (in situ hybridization), or, as described below, it could be localized by virtue of its binding by a radiolabeled ligand (receptor autoradiography). Each of these methods provides a distinct class of information. While immunocytochemistry reflects protein localization, and in situ hybridization can give information about the cell bodies of origin and amount of a specific mRNA, receptor autoradiography reflects the location and amount of binding activity of the receptor protein itself. The central issue in the use of this method is that the binding activity of the protein must be expressed in the tissue preparation under study. It is generally assumed that the receptor protein is intact, is properly inserted in the membrane (if that is its natural location), and is in a chemically proper environment (or at least a workable one). For example, inappropriate salts, fixatives, or protein degradative agents can alter binding. Below we discuss this method in more detail and emphasize some advantages and limitations.

Method

The original work on receptor autoradiography actually began with the infusion of radiolabeled ligands into living animals (21). Under the proper conditions, the ligand passed into the central nervous system (CNS) and was bound, and this ligand–receptor complex could be visualized autoradiographically. This early method was very expensive, time-consuming, and complex. A real revolution in ease, flexibility, and efficiency occurred with the introduction of the tissue slice method (see ) (17, 27, 28, 44). In brief, whole frozen rat brain was sectioned on a freezing microtome (or cryostat) and slide-mounted. These tissue sections (10–30 mm) contained relatively undamaged membranes, including most of the receptor proteins in them. The slides were briefly washed in an appropriate buffer and subsequently immersed in a vial of buffered radiolabeled ligand, during which the receptor–ligand complex forms. (Often the investigator must invest considerable initial effort in establishing the proper binding parameters and conditions, including the buffer composition and concentration, time and temperature of incubation, and, of course, the appropriate ligand and its concentration. It can take many carefully planned runs to optimize these conditions.) Following binding, the slides are usually washed in a cooled buffer to remove unbound ligand (nonspecific binding), thereby leaving specific binding on the tissue section. Again, time, temperature, and the composition of buffer involved in this step are often carefully studied to establish optimal conditions. Finally the slides are removed from the wash and rapidly dried (in a stream of air). The goal here is to remove water quickly to prevent diffusion of the ligand away from the binding site, thus preserving the anatomical precision of the binding.

The binding step itself is often modified for a variety of technical purposes. For example, adding excess unlabeled ligand (or a pharmacological analog) during the binding step is a method for addressing pharmacological specificity and pharmacokinetic issues (Kd, Bmax, etc.). It is often at this step that receptor subtypes and ligand "crossreactivity" issues can be clarified, in that different ligands and conditions can be compared between serial or sequential sections containing the same anatomical structures and receptor composition.

After drying, the slides can be placed next to x-ray film in a light-tight cassette for a period of time ranging from a few hours to months (see ). The range of exposure time of the tissue sections against the x-ray film is determined by the specific activity of the ligand, the number of bound receptors in the section, and the nature of the radioactive marker attached to the ligand. In order to produce higher-resolution information, slides can be fixed and subsequently dipped in photographic emulsion and later developed (). This thin coating of radiation-sensitive particles is capable of providing 1- to 2-mm resolution (in the case of a low-energy 3H emitter). Thus, even intracellular resolution is possible.

Both methods (x-ray film and liquid emulsion) require that the investigator determine the appropriate exposure period. Emulsions, either liquid or film, have a range 10- to 15-fold from threshold detection to complete saturation. For example, if one had a signal 50-fold that of the threshold after 2 days of exposure, the autoradiogram would exceed saturation and be overexposed. For the investigator to attempt reasonable quantitation, it is essential to keep the exposure time within the linear range of the film (roughly the middle 10-fold from threshold to saturation).

Thus, in the example above (50-fold signal at 2 days), one would need to decrease the exposure intensity (from 50-fold to about 5-fold), which could be done by reducing the exposure time by a factor of 10 (to 4.8 hours). Of course, the labeled tissue may be re-exposed to different x-ray films until a correct exposure is obtained. For emulsion-dipped slides, several series of test slides are developed after different time periods to establish optimal exposure time for the body of the experiment. The photomicrograph in illustrates the results of such an autoradiographic study, in this case involving the localization of a class of opiate receptors.

Having established a set of binding conditions for a particular ligand and also having produced autoradiograms within the dynamic range of the emulsion, it is now possible to quantitate the binding. Such quantitation is usually undertaken to compare specific binding across anatomical regions, to evaluate receptor subtypes within a tissue region, or to study changes in receptor levels within regions after various treatments. The methods commonly used in quantitation of autoradiographic data involve digitization of the x-ray (or emulsion dipped) signal, and passage of data to a common personal computer. Along with the digitized value of the autoradiographic signal, separate values for the background area, and area containing only nonspecific binding are usually taken. Through subtraction of background and nonspecific binding, an estimate value for specific binding is obtained. By comparing the signal density (per square micron) with densities produced by a series of known radioactive standards, and knowing the specific activity of the ligand, it is possible to calculate the number of moles of labeled ligand per gram of tissue.

The advantages of receptor autoradiography have become increasingly clear over the last decade, making it a standard technique in a large part of biology. The validity of the method has been supported by data using alternative methods for localization of receptors (or their mRNAs) such as immunocytochemistry and in situ hybridization (see below and Dopamine Receptor Expression in the Central Nervous System). Advantages include:

Limitations

Immunocytochemistry is a method for studying proteins in an anatomical context through the use of specific antibodies directed to detect a particular antigen in tissue (see ref. 34). Briefly, the tissue to be studied is fixed so that the antigen (protein) is kept in place (and not degraded). An antibody directed against that target protein binds it in a thinly sectioned piece of tissue. After washing to remove excess antibody, a second series of antibodies or other marker protein is used to visualize the original (or primary) antibody, thereby indirectly marking the location of the protein in question. As applied to neuroanatomy, this protein might be a neurotransmitter molecule, neurotransmitter-synthesizing enzyme, receptor, or other protein. In addition to neuronal localization, the technique is also widely used to provide valuable information on glia and other non-neuronal cells.

Prior to addressing the method itself, it is valuable to review some basic aspects of the structure of immunoglobulins (IgG in particular) and the antibody–antigen interaction. Antibodies are divided into five classes, with immunoglobulin G (IgG) type being the most commonly used in immunochemical methods. IgGs are 160-kD proteins comprised of two heavy chains and two light chains connected by disulfide bonds (). The structure of the IgG molecule is divided into three domains: hypervariable, variable, and constant. The hypervariable region is the primary source of antigen-binding specificity. Most of the flexibility of IgG binding is genomically determined in this region. The variable region contributes additional sources of protein variation in the unique binding qualities of any particular IgG clone. Finally, the constant region is a nonvariable "structural" component of IgG. This region is common to all IgGs in a particular species. It is known that the hypervariable and variable part of the molecule are the receptor portions of the IgG molecule. Fragments of IgG which contain these regions continue to exhibit active binding. In contrast, the constant region of the IgG molecule can itself serve as an antigen. Thus, in an immunocytochemical reaction the very same IgG may bind antigen (at its hypervariable segment) while its constant region acts an antigen for a second antibody ( and ). Hence, a series of antibodies can be layered, and this amplification can be used to enhance signal strength (35).

An additional set of concepts, that regarding antigen and epitope, warrant discussion. By definition, an antigen is a protein (usually) which can be bound by a specific IgG. An epitope is the few amino acids in the antigen protein to which a specific IgG binds. illustrates the protein sequence with several epitopes (they are often "bends" in the protein. One epitope sequence is highlighted (ABCDE) to indicate the small size of the average epitope (about 4–6 amino acids). Any one IgG molecule (from a single B lymphocyte which itself has been grown in a clonal fashion) binds to only one epitope. In the case of a monoclonal antibody, a single species of specific IgG molecules is used that is specific for a single epitope. A polyclonal antiserum contains multiple IgG species which may bind to any of a number of available epitopes. Interestingly, a monoclonal antibody may be very precise, but because the average epitope is small compared to the total size of the protein (e.g., 4–6 out of 400), one must be concerned about the potential for "cross-reactivity" with an identical sequence in another protein. A polyclonal antiserum reacts with any of several epitopes on the target protein, but the diversity of IgG types may cause cross-reactivity with similar epitopes on other molecules. Overcoming the problem of cross-reactivity often requires fairly thoughtful strategies: multiple antisera, biochemical extraction and characterization of antigen, affinity purification of antibody, and so on.

While the production of antisera and preparation of pure proteins are beyond the scope of this chapter, a few points are worth mentioning. The use of an impure protein preparation (containing multiple proteins) to stimulate antibody production will most likely generate an antiserum with IgGs directed against the protein of interest as well as against protein contaminants. Such a serum may bind some, many, or most of the proteins in an antigen mix. Furthermore, if a large excess of this same antigen mixture is incubated with the antiserum (before it is used on the tissue section), as is commonly performed as a control, the specific signal will probably be blocked and the remaining staining inferred to be specific! Thus, use of a dirty antigen for antibody production can be a source of a mixed antiserum and improper controls.

Assuming that one has a clean antigen and a good antiserum, some considerable effort is then put into establishing optimal conditions for the procedure. Probably the most troublesome is the need to establish fixation chemistry conditions. There is a complex list of fixative reagents, buffers, mixtures, and times which can be tested. Generally, most investigators begin with a neutral buffered 4% paraformaldehyde solution, and then vary mixtures of additional reagents (glutaraldehyde, acroline, picric acid, alcohols, etc.). It is possible to use carbohydrate or lipid fixatives, or even nucleic acid fixatives. The goal is to preserve the antigen in its original cellular context and prevent degradation, while still rendering it "visible" to the antiserum. It is worth noting that cross-linking proteins with fixatives may hide epitopes normally available in the native state; thus, excessive fixation is also a consideration. Once a fixation condition is established, a "working titer" of the antibody is determined. Generally this refers to the dilution of the original antiserum that produces optimal signal while minimizing nonspecific background. Most good antisera can be used in the dilution range of 1:500 to 1:50,000 or more! Other variables include antibody buffer conditions, tissue thickness, and length of incubation. Finally, a detection system (peroxidase histochemistry, fluorescence, etc.) is chosen based on individual needs and level of analysis. This step may also require some fine-tuning.

In the preceding section we have focused on a variety of concepts and issues related to immunocytochemical methods. Here we attempt to present some more detailed considerations concerning application of the technique. The example chosen involves peroxidase histochemistry for visualization, although many of the principles apply to other detection methods as well.

Prior to sectioning, the tissue of interest is usually treated with a protein fixative (e.g., paraformaldehyde). As noted above, the fixation and associated parameters are usually determined empirically. The next step is to dilute the primary antiserum and apply it to the section for 16–48 hours. For the purposes of demonstration, let us begin with a rabbit anti-enkephalin IgG applied to a slice of rat caudate. After incubation with this primary antiserum, the tissue sections are thoroughly washed and a second antiserum is applied for a period of few hours, up to 24 hours. This secondary IgG was produced in goat and raised against rabbit IgG (remember, the primary IgG protein acts as antibody when binding enkephalin, but it acts as antigen when bound by the goat anti-rabbit IgG). We thus have a section with enkephalin bound by a rabbit IgG, which itself is bound by goat IgG directed against rabbit IgG. Following another wash, the next step is the addition of another rabbit IgG for 1–2 hours. This antibody is directed against the enzyme horseradish peroxidase (HRP), but is bound at its constant domain by the remaining binding site of the secondary IgG (goat anti-rabbit). [Note: All IgGs have two binding sites and can thus capture two molecules of antigen.] Following another wash, the tissue is incubated for 1–2 hours in a solution containing the enzyme HRP. [Note: the latter two steps can be replaced with the peroxidase–antiperoxidase (PAP) complex—see ref. 35]. Finally, the tissue is reacted with the HRP substrate hydrogen peroxide in the presence of a chromagen (e.g., diaminobenzidine), resulting in a colored precipitate at the site of this whole antigen–antibody complex. After washing and dehydrating, this precipitate is visible as a brown stain within the antigen-containing structures. The general method can be used at a very high level of resolution (including ultrastructural analysis) and across a large number of proteins (and nonproteins as well). This immunocytochemical sequence, as well as a related strategy (18) involving biotinylated IgGs and the avidin–biotin peroxidase complex (ABC), is outlined in , and an example of its application is shown in .

Technical controls are a very central part of this method. At a basic level it is important to know that the staining seen after the reaction sequence is specific to a particular protein. Criteria that need to be satisfied include the following: (a) absence of staining when the protocol is run with deletion of the primary antibody, (b) absence of staining when antisera is applied that was preincubated in the presence of a large excess of antigen, and (c) lack of staining with application of the enzyme or substrate alone. (Interestingly, some tissues have endogenous peroxidase activity that can be confused with specific staining following immunocytochemical protocols involving this enzyme; this problem can usually be avoided by pretreatment with hydrogen peroxide.) In addition, some non-antigen-related signals are considered to be false-positives. It is important to realize that a "nonimmune" rabbit serum (often used to block nonspecific IgG sites in tissues) is actually loaded with IgGs produced by the rabbit over time, and can therefore be the source of nonspecific labeling. This potential pitfall can be overcome through the use of blocking agents not derived from normal serum (e.g., addition to the incubation mix of carrageenan, a gum that is effective at blocking nonspecific IgG binding sites in tissues). Perhaps the most elaborate control for specificity is the extraction of antigen from tissue, followed by biochemical characterization and quantitation. Other approaches to specificity and precision may be seen in the use of multiple antisera against the same antigen, or even antisera directed against different epitopes on the same antigen.

There are several other considerations that can make immunocytochemistry a complex, expensive, and often technically demanding procedure. Adequate amounts of antigen for blocking studies are often hard to produce, particularly when the antigen is a protein. Even the vaunted specificity of monoclonal antibodies might be defeated by the occurrence of a common epitope in several different antigens. For example, the first four amino acids of all of the opioid peptides are tyrosine-glycine-glycine-phenylalanine. This sequence is found once in the b-endorphin/adrenocorticotropic hormone (ACTH) precursor, three times in the pro-dynorphin precursor, and seven times in the pro-enkephalin precursor! A monoclonal antibody directed against this epitope could potentially find 11 targets in any of three propeptides (see ref. 23). The use of polyclonal antisera can also be somewhat problematic, because they are produced in limited amounts. Moreover, polyclonal antisera can react with different epitopes over different bleeds from the same source.

In situ hybridization is a method which allows for the detection of mRNA molecules, usually within their cells of origin (10, 11, 19, 30, 32, 37, 38, 43). In a sense the method is similar to receptor autoradiography and immunocytochemistry, in that all three techniques rely on some form of binding to form stable complexes. Receptor autoradiography involves binding of receptor by radiolabeled ligand, and immunocytochemistry involves binding of antigen by antibodies. In in situ hybridization, binding occurs between mRNA in the cytosol and an externally produced radiolabeled RNA or DNA probe capable of forming a hybrid with it. Detection of the radiolabeled probe is then performed by methods similar to those used in receptor autoradiography (x-ray film or photographic emulsion).

The most important concept for understanding this technique is that of "hybridization." Two strands of RNA, two strands of DNA, or one RNA and one DNA can bind (or hybridize) through specific hydrogen bonding between chains. In , two strands of RNA are depicted (see inset box). One (AUGCCUCAU) represents a short strand of mRNA; the strand below it (UACGGAGUA) is complementary to it (called cRNA). Inspection of the figure shows that A always binds U and that G always binds C. In fact, A shares two hydrogen bonds with U (or T in a DNA strand), and G shares three hydrogen bonds with C. While any single hydrogen bond is weak, a series of them can form a quite stable complex. In the example shown in there are 5 A-U pairs (10 hydrogen bonds) and 4 G-C pairs (12 hydrogen bonds), resulting in a fairly stable hybrid of 22 hydrogen bonds. A stable combination of RNA–RNA strands is easily obtainable by as few as 25–30 bases (50–75 hydrogen bonds). This simple piece of chemistry forms the basis for a large number of molecular biological methods (in situ hybridization, Northern analysis, Southern analysis, cDNA library screening, etc.), not to mention its central role in stabilization of the double-stranded DNA of the genome itself!

The in situ hybridization method is described below. It is not very different from many aspects of receptor autoradiography and immunocytochemistry. In fact, some aspects of its early development were borrowed directly from these methods. We will center our discussion largely to the application of longer cRNA probes (see ), although it should be noted that the method can be applied with some variation for cDNA and oligonucleotide probes as well.

In situ hybridization is usually carried out on brain tissue that has been previously sectioned and immersed in a formaldehyde fixative. It may also be performed on tissue that has been fixed by perfusion and subsequently sectioned, although this is usually at the expense of some sensitivity. It is important to realize that mRNA molecules are not themselves fixed by a protein fixative such as formaldehyde, but are essentially immobilized by the fixed proteins. Without protein fixation, mRNA can be subjected to the ubiquitous degradative enzyme RNase and/or diffuse away from the tissue section into the buffer. After washing and drying, the tissue is exposed to reagents designed to make it porous to radiolabeled probe. The most common permeabilizer is proteinase K, an enzyme which hydrolyzes many proteins. The tissue is then treated to reduce its net charge (acetylation), which reduces the tendency of the probe to stick to tissues nonspecifically. After another wash step, the sections are incubated with radiolabeled cRNA probe for 12–48 hours. During this period, the cRNA probe (which is applied in great excess) can form stable hybrids with endogenous mRNA. It is this RNA–RNA hybrid which provides the basis for signal localization, as well as a series of specificity controls. Following hybridization the tissue is incubated in the enzyme RNase, which digests all single-stranded RNA and thus digests nonhybridized probe, thereby dramatically reducing a prominent source of background. After the RNase step, the sections are washed in low salt buffers at 45–65°C. This important step removes nonspecifically adherent probe, as well as probe that is only weakly hybridized. The tissue section is then dried, and subsequently it is exposed to x-ray film or is dipped in photographic emulsion. Quantitation is performed almost identically to receptor autoradiography (see previous section). An example of the results of a typical in situ hybridization experiment is illustrated in .

A number of technical issues are important in understanding in situ hybridization technology. The preceding description was centered around the use of longer cRNA probes (approximately 200–1000 bases) or riboprobes, in which labeled nucleotides are incorporated into the RNA strand. An alternative method is to use short cDNA probes (generally 20+ bases) to hybridize to mRNA. These oligonucleotide probes have some advantages, particularly concerning tissue penetration, which is facilitated as a result of their small size. They are also easier to use than cRNA probes, but cannot be labeled to an equivalent high specific activity. This loss of sensitivity can be overcome through the use of multiple nonoverlapping oligonucleotide probes. It is also notable that in addition to an isotope of sulfur (35S) as label, other isotopes are available for use, such as 32P and 33P. 32P is a very-high-energy label with wide scatter, and it generally produces a less desirable signal compared to 35S, which is only half the specific activity of 32P but offers 30–50 times higher resolution. 33P offers a resolution and specific activity that is intermediate, though closer to that of 35S than 32P. The shorter half-life of 33P makes it attractive relative to 35S in terms of safety and disposal, although it presently is more expensive than either 35S or 32P. It should be noted that nonradioactive markers are being increasingly used to produce a high-resolution signal for in situ hybridization (reviewed in refs. 30 and 43). These materials are usually simple molecules which are coupled to a nucleotide and which, once incorporated into probe and applied to tissue, can be detected with methods involving or resembling immunocytochemistry. While presenting an obvious appeal in terms of rapid detection, ease, and safety of use, these methods are typically not as sensitive as their radiolabeled counterparts (30,43). However, in some cases, immunocytochemical detection of the nucleotide-coupled molecule can provide further signal amplification (22).  Also, a number of protocols have recently been advanced for the simultaneous detection of two mRNA species in brain by combining radioactive and nonradioactive detection methods (see ) (25, 26, 42). The in situ hybridization technique may also be adapted for studies at the ultrastructural level, particularly those methods involving histochemical detection of nonisotopic probes (40).

Another important issue concerns specificity controls. Probes are capable of "cross-hybridization" in a fashion analogous to the cross-reactivity of antibodies. In the event that an investigator suspects cross-reactivity, the first and conceptually simplest control for this phenomenon is degradation (using RNase) of all RNA in the section prior to hybridization, which should remove the signal. A second control is the use of a "sense" strand RNA probe, which can be created by transcribing the opposite strand of the gene (opposite to that used to create the usual cRNA or "antisense" probe). This radiolabeled sense probe should not yield a signal, because it is actually identical to a part of the mRNA sequence. Assuming that the RNase and sense strand controls are negative, the investigator may then want to establish a very stringent set of temperature and salt conditions for the wash step of the mRNA–cRNA hybrid. Mismatched strands giving rise to a false signal can be melted apart with increasing temperature and decreasing salt conditions. Finally, of course, mRNA localization should be compared to known data concerning protein expression or localization.

The method, in its most sensitive form (use of riboprobes), is complex, expensive, and requires skills in molecular biology.

The methods presented above (receptor autoradiography, immunocytochemistry, and in situ hybridization) involve the study of biochemical processes and structures. In contrast, neuroanatomical tract-tracing is performed to answer fundamental questions concerning neuronal projections and connectivity. The past two decades have seen great advances in the development of new and highly sensitive methods for tracing neuronal connections, and even a superficial survey of all of these techniques is well beyond the scope of the present chapter. In general, many modern neuroanatomical tracing methods involve the injection of various types of compounds or dyes that are subsequently transported in an anterograde or retrograde fashion (or both) in relation to neuronal cell bodies. These compounds may be visible through fluorescence microscopy or may be detected using standard histochemical or immunocytochemical techniques. Through the placement of discrete injections within a given brain region, it is thus possible to confirm the inputs and outputs of an area. Multiple compounds can be injected in the same animal to address issues such as axonal collateralization (e.g., two or more retrograde compounds) (1, 29, 41) or to establish neuronal chains (through the use of multiple tracers) (7, 8). The capacity to visualize many of these compounds at the ultrastructural level has added a new level of power in defining synaptic connectivity (3, 13, 14, 24, 45). Moreover, many of these methods are compatible with immunocytochemistry (29, 33) or with in situ hybridization (6, 39), making it possible to reveal not only the presence of a given projection, but also its biochemical composition (e.g., neurotransmitter content) in the same experiment. A sampling of some of these applications is illustrated in .

At the technical level, a number of issues require consideration (for discussion, see refs. 2, 15, 16, and 24). For example, it is obviously desirable to limit the extent of tracer diffusion to the specific anatomical structure under study, because spread outside the target area confounds interpretation. In many cases this can be accomplished by iontophoretic delivery or, alternatively, by limiting injection volumes. Individual characteristics of the tracer compound must also be considered, such as tendency to label fibers-of-passage, toxicity, diffusibility, and tendency for transport in the direction opposite to that desired. Fixation parameters are also of importance, particularly for experiments requiring immunocytochemical detection or those that will ultimately rely on electron microscopic evaluation.

Let us briefly return to our example of the dopamine neurons in the substantia nigra presented in the early part of this chapter. As part of the classic nigrostriatal pathway, this projection (as well as nonstriatal projections of these neurons) could be demonstrated using an anterograde tracer (e.g., PHA-L, biocytin, Fluoro-Ruby). These cells also could be retrogradely labeled from the striatum with any of a number of compounds (e.g., Fluoro-gold, true blue, rhodamine beads, WGA-HRP, or a cholera toxin B-HRP conjugate) and in a combined immunocytochemical experiment shown to contain dopamine (or one of its synthesizing enzymes, such as tyrosine hydroxylase). Such retrogradely labeled cells could also be shown to express a neuropeptide transmitter or a specific dopamine receptor subtype (or their respective mRNA transcripts) in an experiment combining retrograde tracing with immunocytochemistry (or in situ hybridization). Of course, as with any combination of methods, a set of technical issues and limitations concerning compatibility must be considered in the design of the experiment. Nevertheless, recent advances in such combined applications have made it possible to characterize brain regions in substantial detail with a high degree of flexibility (3, 4, 6, 7, 8, 9, 14, 20, 29, 33, 39, 45). For illustrative purposes, brief descriptions of an anterograde tracing method (PHA-L) and a retrograde tracing technique (Fluoro-gold) presently in common use are included below.

Anterograde tract-tracing using the lectin Phaseolus vulgaris-leucoagglutinin (PHA-L) was first introduced in 1984 by C. Gerfen and P. Sawchenko (12). This protein is delivered by iontophoresis, incorporated largely by dendrites at the site of injection, and is subsequently transported along the axon (at a relatively slow rate of approximately 4–6 mm/day). Thus, it is possible to label all neurons giving rise to a projection (including substantial portions of their dendritic arbors, in many cases resembling preparations using the Golgi method), as well as produce a rather complete picture of axonal trajectories and terminal arborizations. Importantly, when delivered under the appropriate conditions (iontophoresis of a 2.5% solution), this tracer is apparently neither incorporated by passing fibers, nor transported retrogradely by terminals in the vicinity of the injection site (12). Following survival periods, perfusion fixation (usually with a buffered formaldehyde or paraformaldehyde solution) is performed and brains removed and subsequently sectioned. The lectin is then detected in tissue sections using standard immunocytochemical methods (typically involving the avidin– biotin method as outlined in , rendering the technique compatible with a wide array of immunocytochemical procedures. For example, neurochemical identification of PHA-L labeled axons may be performed with double-immunofluorescence methods, and biochemical characterization of the targets of PHA-L-labeled afferents may be accomplished in combination with immunocytochemical techniques (see ). The method may also be extended to analyses at the electron microscopic level, in which synaptic interconnections can be unequivocally demonstrated (13, 14, 45). A disadvantage of the method is the somewhat limited penetration of the anti-PHA-L antibody. This can be particularly troublesome for analysis of synaptic connectivity in electron microscopic studies that require the omission of the membrane-solubilizing detergents in order to preserve ultrastructure.

Perhaps one of the most versatile fluorescent compounds available for retrograde tract-tracing is Fluoro-gold. Fluoro-gold is a stilbene derivative first introduced in the mid-1980s by L. Schmued and J. Fallon (31). Although the mechanism of incorporation is unknown, Fluoro-gold is transported relatively rapidly in the retrograde direction and produces visible labeling within neurons within 24–48 hours. Longer survival times (10–14 days) typically produce a stronger fluorescence signal, including extensive visualization of dendrites. The tracer is compatible with many standard fixation protocols and is amenable to numerous combinations, including those involving multiple fluorescent markers, in situ hybridization histochemistry, and immunocytochemical methods. Additionally, the recent development of an antibody directed against Fluoro-gold (5) has made it possible to detect the tracer immunocytochemically (see ), allowing important advantages in terms of increased sensitivity, enhanced resolution (particularly of dendritic processes), and compatibility with ultrastructural analyses. A limitation of the technique is its apparent ability to be taken up by fibers-of-passage.

In summary, the advent of a new generation of sensitive and versatile tract-tracing methods has helped make it possible to define neuronal circuits with a new level of sophistication and precision. We have highlighted only two of these tools in an effort to illustrate their suitability for combined studies; these obviously represent only a small fraction of the tracing compounds currently available for neuroanatomical investigation. A number of excellent texts are available for detailed descriptions of modern neuroanatomical methods, particularly with respect combined applications (see refs. 2, 15, 16, and 24)

The study of brain in cellular and circuit contexts has relied heavily on the methods overviewed in the present chapter. Each of these techniques (receptor autoradiography, immunocytochemistry, in situ hybridization histochemistry, and neuroanatomical tract-tracing) can be used to provide fundamentally different classes of information concerning specific neuronal elements. We have attempted to highlight some of the basic principles underlying these techniques. While each method presents its own set of advantages and limitations, both individually and in concert with other techniques, new and creative combinations continue to emerge that have enabled characterization of neuronal microcircuitry in ever-increasing detail. The ultimate power of these approaches lies in the ability to provide functional insights by the coupling of information gained from them. Such circuit analyses are contributing substantially to our understanding of fundamental mechanisms underlying brain function in health and disease states.

We wish to thank Dr. Derek T. Chalmers and Dr. Charles A. Fox for helpful comments regarding the manuscript. We would also like to thank the following for photomicrograph contributions: Drs. Eileen J. Curran, Charles A. Fox, Alfred Mansour, Paul E. Sawchenko, Larry W. Swanson, and László. Záborszky.

published 2000