Additional related information may be found at:
Neuropsychopharmacology: The Fifth Generation of Progress

Back to Psychopharmacology - The Fourth Generation of Progress


Beyond Second Messengers

James I. Morgan and Thomas E. Curran


The interaction of extracellular ligands with their receptors on the plasma membrane elicits the flow of information into the cytoplasm via numerous signal transduction pathways. Historically, this information flow was considered to bring about rapid alterations in cellular functions by modifying the activity of existing proteins. However, it is now clear that signal transduction cascades do not terminate in the cytoplasm but rather extend to the nucleus where they are capable of bringing about alterations in gene expression (Fig. 1). This provides a mechanism whereby a cell can adapt to alterations in the extracellular milieu by changing the levels, or patterns, of gene expression (see Basic Concepts and Techniques of Molecular Genetics).

Oncogenes are defined as any genetic element that can cause cellular transformation, and proto-oncogenes are the normal cellular genes from which the oncogenes are derived. Historically, oncogenes were first described as the transforming elements present in some retroviruses. Therefore, the viral oncogene was designated v-onc to distinguish it from its normal cellular counterpart (the proto-oncogene), c-onc. By convention, most genes or their messenger ribonucleic acids (mRNAs) are referred to by three letter names in lower case that are italicized, whereas their protein products are not italicized and use upper case. For example, v-src is the viral oncogene (or its mRNA) that encodes the protein v-Src. Proto-oncogenes can be activated to become oncogenes by many other mechanisms besides capture by retroviruses (which is actually relatively rare). Such processes include mutation, rearrangement, or amplification (reviewed in ref. 34). In these cases, the c- prefix is not used, because no viral homolog exists. Many proto-oncogenes encode proteins that fulfill a role in signal transduction pathways; these functions include extracellular growth factors, membrane receptors, cytoplasmic and membrane-associated protein kinases, guanosine triphosphate (GTP) -binding proteins, and transcription factors. A generic signaling pathway is depicted in Fig. 1 that shows the flow of information from the extracellular milieu to the cytoplasm and finally into the nucleus. In this cascade of events, there is a further distinction between events that require a nuclear response (long-term cellular responses) and those that are protein-synthesis independent (short-term cellular responses). Examples of specific proto-oncogenes are given within the framework of this general signal transduction pathway to illustrate the range of their biochemical and cellular properties (see Signal Transduction Pathways for Catecholamine, The Neurobiology of Neurotensin, and Intracellular Messenger Pathways as Mediators of Neural Plasticity).

A more detailed list of proto-oncogenes with their functions (where known) is provided in Table 1. Proto-oncogenes can be conveniently classified by the role they play in signal transduction. In Table 1, we have arranged specific proto-oncogenes into various general classes such as receptor ligands and membrane receptor tyrosine kinases. The order of presentation is designed to indicate the stage in the signal transduction pathway that particular proto-oncogenes act starting in the extracellular milieu, moving to the membrane, through the cytoplasm, and into the nucleus. Where it is known, we have indicated the specific function of particular proto-oncogenes. For example, the proto-oncogene c-sis encodes the B chain of platelet-derived growth factor. In many cases, proto-oncogenes belong to a gene family in which the function of one member is known. In this case we refer to the proto-oncogene function as -like. For example, the proto-oncogene int-2, encodes a protein that is related to basic fibroblast growth factor.

As our knowledge of stimulus釦ranscription coupling has grown it has begun to reveal the molecular basis of a number of neurobiological processes. Some of these processes were already known to be protein-synthesis dependent and, therefore, likely to have a transcriptional component. However, in other instances the finding of rapid alterations in gene expression was quite unexpected and has lead to a reexamination of the molecular basis of these responses. From the neuropharmacological perspective, two types of studies involving gene expression in the nervous system are relevant. First, monitoring gene expression can provide insights into which cells are involved in particular neurobiological or neuropathological responses. Second, the molecules that couple second messengers to gene expression, as well as the products of these genes, offer new substrates for potential neuropharmacological intervention. Although there is considerable interest in developing therapeutics that act on gene transcription, this is a field still in its infancy. Therefore, this chapter will deal with methods of studying neurobiological and neuropharmacological responses by assessing immediate-early gene (IEG) expression.

Much of what we know concerning stimulus釦ranscription coupling in the nervous system has its origins in studies of cancer biology. One critical discovery was the recognition that most proto-oncogenes, the normal cellular homologs of oncogenes, encoded proteins involved in signal transduction (see Fig. 1 for details) (34). (Oncogenes are defined as being any genetic element that can cause cellular transformation.) Furthermore, and somewhat surprisingly at the time, some proto-oncogenes (e.g., erbA, c-myb, c-myc, and c-fos) encoded nuclear proteins that, with the exception of erbA (the thyroid hormone receptor), did not fit into any conventional signal transduction pathway. It was supposed that the products of these genes should act as transcription factors, although little direct evidence existed for this assumption at the time. Subsequently it was discovered that addition of serum to quiescent fibroblasts resulted in a rapid and transient induction of several of these proto-oncogenes (4, 12, 27, 36), the best known being c-fos, c-jun, and c-myc. Therefore, it was postulated that these inducible proto-oncogenes might encode transcription factors that could act as nuclear "third messengers," coupling second messenger-mediated events to subsequent alterations in gene expression (5, 33). In the meantime a considerable body of evidence has accumulated to support this view, although the entire picture is still far from clear.

Although rapidly inducible proto-oncogenes, such as fos, were studied primarily in the context of mitogenesis and transformation, it soon became clear that they could be induced by stimuli that had no influence on cell proliferation (6, 11). Indeed, they could be activated by depolarization of neurons (32), cells that are not even competent for mitosis. Therefore, it was thought that the signal transduction pathways utilizing inducible proto-oncogenes should be involved in many biological processes and would be used as ubiquitously as any of the more conventional second-messenger systems (5, 33).

In parallel with the quest for the function of proto-oncogenes, other studies were designed to identify genes that were activated by serum treatment of quiescent fibroblasts. These investigations identified a range of inducible genes; some were proto-oncogenes, others were structurally related to known proto-oncogenes (i.e., members of gene families such as fosB and junB) yet others had no obvious homology to known proto-oncogenes and encoded proteins that were either of unknown function or that were involved in signal transduction (e.g., tyrosine phosphatases and cytokinelike molecules). Therefore, the view arose of a much more global transcriptional response to stimulation that involved whole families of genes whose products might contribute to signal transduction both within and between cells.

It was subsequently found that many of the genes that were activated immediately after serum addition were transcriptionally induced even if protein synthesis was blocked. Indeed, they were superinduced in the presence of agents such as cycloheximide and anisomycin. This unifying property lead to the definition of these genes as constituting the cellular IEG class (5, 26). This name was adopted from the IEGs of viruses, which are expressed promptly upon infection of a host cell even in the absence of protein synthesis. The property of protein synthesis independence is taken to mean that all of the signaling and regulatory molecules necessary to elicit the transcriptional response of this class of genes are already present in the unstimulated cell. Furthermore, since the function of IEGs of viruses is to control the expression of early genes and progression of the viral replication cycle, so the cellular IEGs (cIEGs) were postulated to control the expression of genes necessary for subsequent adaptive programs. As noted above, not all cIEGs encode transcription factors. Thus the cellular immediate-early response should be viewed as an integrated biological process involving transcriptional changes and modification of signal transduction pathways in the stimulated cell and dissemination of the response to adjacent cells and tissues (Fig. 2) (for reviews see refs. 5, 17, 30, 31 and 43).

Cellular responses to extracellular stimuli can be divided into those that are dependent upon protein synthesis and those that are not. In general, transcription independent responses tend to be rapid events that are mediated by second messenger molecules that control the posttranslational modification of preexisting substrates such as ion channels, membrane receptors, and cytoskeletal proteins. Such changes result in alterations in the properties of the substrates that might include alterations in ion gating or changes in secretion of neurotransmitters. This type of process is referred to as the short-term response to stimulation. The cell also mounts a delayed, but more protracted, response to stimulation that is referred to as the long-term response. This involves second-messenger molecules altering gene expression. The example shown in Fig. 2 focuses on the immediate-early genes, although second messengers (and some ligands such as retinoids and steroids) can act to alter expression of many other classes of genes. Immediate-early genes frequently encode transcription factors that control the expression of further, so-called target, genes. In the hypothetical scheme in Fig. 2, it is supposed that target genes might be those encoding neurotransmitter synthesizing enzymes and receptors, neuropeptide precursors, and ion channels. In addition, some immediate-early genes encode secreted proteins or enzymes involved in second-messenger metabolism. These can be viewed as molecules that disseminate the response to other cells and alter the coupling efficiency of the signal transduction pathway. It should be emphasized, however, that this scheme is hypothetical and grossly oversimplified. No target genes for particular cellular immediate-early gene products have been identified in neurons, and it is most unlikely that the regulation of any individual gene is controlled by a single transcription factor or class of factors, but rather involves complex combinatorial interactions between resident and inducible proteins.

Fos, Jun, and Activator Protein-1

Many of the applications of cIEG expression in neuropharmacological investigations stem from the properties of two prototypic members of this gene class, c-fos and c-jun. Both of these inducible proto-oncogenes encode nuclear phosphoproteins (Fos and Jun) that physically associate with one another through an amphipathic a-helical domain containing a heptad repeat of leucine residues that has been termed a leucine-zipper (25). Both Fos and Jun, as well as all other members of the fos and jun gene families (i.e., fosB, fra1, fra2, junB, and junD), possess leucine zippers. All members of the fos family of proteins can form heterodimers with any member of the jun family. In addition, proteins of the jun family can form hetero- and homodimers among themselves, a property that is not shared by Fos or its family members (Fig. 3) (for reviews see refs. 7 and 23).

Subsequent to dimerization, the various homo- and heterodimeric complexes bind to a specific deoxyribonucleic acid (DNA) sequence, TGACTCA (7, 24, 37). Independently, this sequence was identified as the canonical binding site for the transcription factor activator protein 1 (AP-1) and is essential for both basal and stimulated transcription from several genes (reviewed in ref. 7). Subsequently, AP-1 was shown to be comprised of Fos and Jun as well as several Fos- and Jun-related proteins. Binding of deoxyribonucleic acid is largely confirmed by regions in Fos and Jun that are rich in basic amino acids. Each member of the dimer contributes a half-site for DNA binding. Whereas all dimeric combinations of AP-1 interact with the same, or similar, DNA sequences, it is to be expected that there must be dimer-specific differences in the details of binding and/or the consequences for transcription. Indeed, distinctions can be detected with DNA binding and transactivation assays. For example, Fos褒un heterodimers bend DNA in a different direction to Jun褒un homodimers (23). Such an effect might account for why some combinations of AP-1 dimers promote transcription from a particular target gene whereas others inhibit expression.

Both leucine zipper and basic DNA-binding domains have been detected in other families of transcription factors, such as the activating transcription factors (ATFs) and cyclic adenosine monophosphate response element binding proteins (CREBs). This led to the coining of the term basic-zipper protein to describe the members of a superfamily of transcription factors that includes the Fos and Jun families of proteins as well as CREBs and ATFs (24). Furthermore, various members of this superfamily are capable of cross-family dimerization (15). For example, Jun may dimerize with CREBP1, giving rise to a heterodimer that binds to the cyclic adenosine monophosphate (cAMP) response element (CRE, TGAGCTCA) rather than the AP-1 site (TGACTCA). Thus, the use of the term AP-1 to describe Jun, is a misnomer, since its DNA-binding specificity (and possibly transactivating potential) is a function of its partner in the dimer. One other point should be noted here, namely that many members of the basic-zipper superfamily are not inducible genes. In fact, junD, a member of the jun family, is a constitutively expressed transcription factor. Therefore, a more accurate view of the immediate-early response as it applies to members of the basic-zipper family is one in which there is a varied and complex pattern of dimer formation between both inducible and constitutively expressed transcription factors. Although not extensively investigated, it is supposed that the precise details of this pattern are a function of the cell type and the stimulus used. Likewise, it is postulated that the genomic targets for these dimers and the consequences for transcription will also be dictated to some degree by cell type and stimulus.

Diverse types of extracellular stimuli, via second messenger molecules, elicit the rapid transcriptional activation of cellular IEGs such as c-fos and c-jun. Both c-fos and c-jun belong to gene families whose products in turn share general structural features with other transcription factors that together constitute a superfamily, termed basic-zipper proteins. Further families of basic-zipper proteins include the CREBs (cyclic AMP response element binding proteins), ATFs (activating transcription factors) and the Maf proteins. These proteins all form homo- and heterodimers by way of a leucine-zipper structure, shown in Fig. 3 as four horizontal bars. These dimeric complexes then show relatively specific binding to short DNA elements that is mediated by domains rich in basic amino acids. In the case of Fos褒un heterodimers, DNA binding is to the AP-1 (activator protein-1) consensus site (TGACTCA). The binding of the dimeric complex is then believed to contribute to the transcription of genes bearing such sites in their promoters. Although many members of the fos and jun gene families are IEGs (e.g. fra-1, fosB, junB) others, such as junD, are constitutively expressed as are many members of the basic-zipper superfamily, such as CREBP1 in the example shown. Furthermore, these additional families of basic-zipper proteins can form intra- and interfamily dimers that can interact with other consensus DNA-binding sites. Thus, a more realistic picture of the IEG response as applied to the basic-zipper family is one in which inducible and constitutive proteins interact in a temporally dynamic manner to control transcription of target genes. DNA binding specificity is determined by the precise composition of the complex. In the example shown in Fig. 3, Fos褒un dimers bind at the AP-1 site, whereas Jun-CREBP1 dimers bind at a CRE site. However, it is now known that other consensus sequences exist for other dimer configurations. Obviously this situation provides for enormous diversity, and it is presumed that the phenotype of the cell and the nature of the stimulus must determine the net genomic response.

When fos and jun expression was investigated in the nervous system, a further property of AP-1 dimer formation emerged. Seizures result in a transient induction of c-fos and c-jun messenger ribonucleic acid (mRNA) in the brain (49). However, when gel retardation and immunoblot analyses were carried out on nuclear extracts from brains of mice that had received seizures, a paradox emerged (47). Although fos mRNA and protein appeared and disappeared in the brain within 3 to 4 hr of a seizure, AP-1 DNA-binding activity was elevated for 8 to 17 hr (depending on the type of seizure). This was explained by the fact that there was a delayed, but protracted, induction of several proteins that cross-reacted with the Fos-antiserum that could also participate in AP-1 complexes. That is, there was a staggered appearance and disappearance of Fos and several Fos-related proteins that can all contribute to AP-1-like complexes. Thus, the composition of AP-1 alters in a dynamic and reproducible manner with time after seizure (30, 47). Therefore, when analyzing components of the immediate-early gene response, time is a critical variable.

Immediate-Early Gene Expression in the Nervous System

Even though the study of immediate-early gene expression in the nervous system has burgeoned in recent years, basically only two types of experiments have been performed. The large majority of the studies have involved the monitoring of one or another IEG product (usually Fos) as a surrogate marker of neuronal activation. These investigations range from the analysis of IEG expression in cell culture to measurements of IEG levels and distribution in the brain following the application of many types of physiological, pharmacological, and behavioral stimuli. These studies are now so numerous (in excess of 500) that it would be inappropriate to elaborate them here and the interested reader is directed to a number of reviews on the subject (30, 31, 43). However, specific examples are used to illustrate particular methods or approaches in IEG research. A lesser number of investigators have tried to establish the roles that IEG products might play in particular neurobiological processes by trying to perturb their expression or activity.

Mapping, studies have utilized a range of techniques that include, Northern and Western blotting, immunohistochemistry, in situ hybridization, and gel retardation assays, as well as transgenic animal and DNA transfection technologies. Investigations of IEG function have proven much more difficult and have involved a range of approaches such as in vitro transcription assays, transfection analyses, antisense, and gene knockout methods, as well as transgenic mouse experiments utilizing targeted overexpression of particular IEGs or inhibitors (e.g., transdominant suppressors). Given the preponderance of the mapping type of study, most of the methodological details and critique elaborated in this chapter are focused on this type of analysis. In addition, a novel approach is presented for IEG mapping that employs transgenic fos僕acZ reporter mice. These animals represent a model in which one can rapidly and unambiguously follow IEG expression with single-cell resolution.



Biochemical Analyses

The least ambiguous and simplest measure of IEG expression involves Northern transfer and hybridization (28, 49). This method has the advantage of clearly identifying the transcript, it is quantifiable, and it can be applied to multiple IEGs in the same sample. Its principal limitation is that it lacks cellular resolution and requires a substantial amount of tissue. Although a polymerase chain reaction (PCR) may circumvent the sample size problem, thereby permitting a finer regional localization, this method is notoriously nonquantitative and should, in any case, be confirmed by Northern analysis. In terms of sensitivity, RNAase protection assays lie between Northern blots and PCR, and they can be made quantitative. However, the method is far less routine than the other two and conditions need to be worked out for each IEG transcript. Finally, Northern blots, RNAse protection, and PCR may not necessarily reflect the level of the cognate protein, because mRNA for IEGs can accumulate to high levels in the absence of protein synthesis. This is not a facile point, because a number of neuropathological situations, such as cerebral ischemia, may be associated with a block of translation. This is known to have produced confusing, and sometimes contradictory, results in studies of IEG expression in animal models of stroke (13, 14, 21, 22).

To circumvent the issue of extrapolating alterations in mRNA levels to changes in protein levels, a number of investigators have used immunoblotting to assess IEG expression (e.g., 47). This procedure still has the limitation of lacking cellular resolution, but it has proven useful in establishing the time frame of changes in the expression of some IEGs. However, this approach does have additional complications. First, sample preparation becomes a significant issue, because some authors have found that crude nuclear extracts must be prepared from brain to obtain adequate Western blots for some IEGs. Second, the validity of the data relies solely on the specificity and affinity of the antisera or antibodies used. Because many of the IEGs belong to gene families, some antisera are known to cross-react with related proteins and reveal bands of the incorrect size in immunoblots (47). Unfortunately, few of the investigators using this approach run authentic (i.e., recombinant) protein standards.

A further biochemical approach for investigating IEG expression takes advantage of the fact that many of their products bind to specific DNA sequences. Thus their levels can been determined in a semiquantitative manner by so-called gel retardation or mobility shift assays (47). In this procedure protein extracts are mixed with radioactively labeled double-stranded oligonucleotides that contain the consensus-binding sequence for particular IEG transcription factors. These complexes are then run on a nondenaturing gel, which separates bound from free probe. The bound and unbound species can be detected subsequently by autoradiography. This strategy has proven very useful in analysis of AP-1 levels in both mouse and rat brains following a number of challenges, including seizures and chronic administration of drugs of abuse (see Intracellular Messenger Pathways as Mediators of Neural Plasticity, this volume). In addition, by using other consensus sites, it has proven possible to analyze binding of CREB proteins in brain extracts.

In our hands, gel retardation analysis is most reproducible when using crude nuclear brain extracts rather than whole homogenates. This nuclear preparation has the added advantage that it is amenable to assay by immunoblotting, which can be performed in parallel with the gel shift and provides a further level of information concerning the composition of the complexes (47). The analysis of the components of particular complexes can be extended by the use of commercial antisera to many components of the CREB, ATF, and AP-1 families. These are useful not only for immunoblots but also gel shifts. Conventionally, one pretreats the nuclear extracts for from a few hours up to a day (in some cases) before adding the labeled oligonucleotide and performing the gel shift. Active antisera may either supershift the complex (by contributing to its mass and further slowing the probe's migration) or destroy its DNA binding activity. A further proof of specificity relies on appropriate patterns of displacement of radiolabeled probe binding by unlabeled oligonucleotides. In many cases, investigators have used unlabeled cognate probe and an irrelevant oligonucleotide. However, it is now apparent that such controls for specificity are incomplete. Just because a complex forms on an AP-1 site does not necessarily mean that it will not bind as well or better to other consensus sites. For example, complexes that bind to AP-1 sites often bind just as well to CRE sites. Thus competition curves should be performed for various canonical binding sites and to sequences containing point mutations that destroy biological activity.

One advantage of the gel shift method is that it involves the assay of an intrinsic biological property of a protein(s) and, therefore, requires neither prior knowledge of the composition of the complex nor specific reagents. Thus it nicely complements the other biochemical methods that rely exclusively on hybridization of specific copy DNA (cDNA) probes or reaction with specific antisera. However, like other bulk methods, gel shifts do not provide cellular resolution. Another limitation is that they are notoriously sensitive to the incubation conditions. Indeed, there are a number of fundamentally different sets of incubation conditions for gel shift analysis. For example, some workers use low ionic strength buffers, whereas others use high ionic strength. Another example specifically relates to AP-1 binding. It is known that recombinant Fos and Jun do not bind the AP-1 consensus site with high affinity unless they are in a highly reducing environment; usually 5 mM diothiothreitol (DTT) (1). The molecular basis for this effect has been elucidated and involves the reduction of a single cysteine residue that appears to spontaneously oxidize to an uncharacterized intermediate (e.g., sulfinic acid derivative). These redox-driven changes on cysteine do not involve disulfide bond formation. A cellular protein, Ref-1, has been cloned that is able to reduce the partially oxidized cysteine present in both Fos and Jun to the sulfhydryl form (51). Therefore, details of tissue preparation and reducing capacity are especially critical when determining AP-1-like DNA binding activity. Nevertheless, some studies still employ 1 mM DTT, which is normally insufficient to maintain the cysteine in Fos in its sulfhydryl (DNA-binding) form. Unfortunately, these disparate assay conditions often make it difficult or impossible to directly compare results obtained in different laboratories.

Anatomical Methods

The fundamental limitation of the preceeding methods is that they all lack cellular resolution. That is, they cannot reveal precisely which cells are expressing particular IEG products. Conventionally, two methods have been used to investigate the cellular localization of IEG products: namely, in situ hybridization and immunohistochemistry. These are the anatomical homologs of Northern and Western blots, respectively, and have the same limitations regarding their dependence on the availability of specific reagents. Nevertheless, both of these techniques have been widely applied as neuroanatomical mapping techniques. Indeed, their use has burgeoned with the availability of commercial sources of antisera to Fos and several other IEG products (31).

Initially, immunohistochemistry for Fos was limited by the fact that most antisera were supplied by individual researchers and several of these reagents were known to cross-react with other inducible proteins (10, 19, 29, 41). Indeed, one of the most commonly used antisera revealed a composite picture of Fos-like immunoreactivity (FLI) that included Fos, FosB and two unidentified inducible proteins of approximately 35 and 46 kDa, respectively (47). In the unstimulated brain, the predominant contributor to FLI is an unidentified 35-kDa Fos-related antigen (FRA 35K). During the first 2 hr after application of a convulsant stimulus, the predominant contributor to FLI is Fos itself. Subsequently Fos levels wane, and by far the major components of FLI are the FRA-35K and FRA-46K. The latter component can be detected 18 to 48 hr following certain types of seizures (48).

Fos immunohistochemistry has been used to determine the pattern of activation of neurons in many situations. Administration of many types of neuropharmacologically active substances induce FLI. It would be impossible to document all the examples of these applications, because it would reduce this chapter to little more than a catalog of drugs and effects on FLI. Likewise, the constraint on citations precludes exhaustive referencing. Therefore, in the absence of any current reviews we have listed a number of more recent papers that provide a more detailed literature base. Thus, FLI has been used to investigate the pharmacology of dopamine D1 and D2 receptors (see Dopamine Receptor Expression in the Central Nervous System, this volume), nicotinic (39) and muscarinic (18) cholinergic receptors, serotonin (40) and catecholaminergic (50) systems, as well as N-methyl-D-aspartate (NMDA) and non-NMDA glutamate receptors (2). It has also been used to study more complex neural processes such as long-term potentiation (8), circadian rhythmicity (38), and behavioral phenomena such as kindling (10) and learning (3).

Studies with antisera to the Jun family of proteins are now beginning to reveal interesting patterns of expression in a number of neurobiologically relevant situations (16, 20). Like c-fos, both c-jun and junB are activated by numerous types of stimuli in a characteristic spatial and temporal manner in the brain. Although many of these sites and situations are the same as those in which FLI is detected, some unique patterns of expression have been observed. A notable example is the persistent expression of Jun (but not Fos) following axotomy (16, 20). A further family member, junD, appears to be expressed continuously and at relatively high levels in the brain. Although junD is not an IEG, sensu stricto (it is not superinduced by cycloheximide treatment), some reports have indicated localized and delayed increases in its expression in the nervous system (e.g., 16). However, in our hands, expression of junD mRNA is constant in several seizure models in rats and mice.

Published protocols for immunohistochemistry are diverse, although generally the most reproducible and convincing results have been obtained on floating cryostat sections. Many of the references cited above for the application of Fos mapping in the brain contain details of successful immunohistochemical protocols. In most all instances, FLI is nuclear and one observes exclusion from the nucleolus and its associated heterochromatin (35). A number of studies have shown cytoplasmic staining using anti-Fos antisera, although the only case for which there is any evidence of Fos being present outside of the nucleus involves cell death. Thus, when showing cytoplasmic FLI, an independent confirmation is required (e.g., in situ hybridization or a biochemical characterization), and one should consider the likelihood that a cross-reacting material is being detected.

The major limitation of in situ hybridization is not so much specificity (although this may be a concern with closely related gene products or where nucleotide sequences are based upon cDNAs from other species) but rather the fact that it does not measure directly the level of IEG proteins. In addition, there may be problems with detection limits and quantitation. Like immunohistochemistry, this approach has been quite rewarding and is especially useful when combined with immunohistochemistry for IEG products or other proteins (e.g., neuropeptides or cell-specific markers).

Because immunohistochemistry and in situ hybridization are methods that are generally slow and prone to ambiguity, we have developed a completely novel approach to neuroanatomical mapping of IEG products. The strategy has been to introduce a readily detectable reporter gene (bacterial b-galactosidase) into the IEG of interest in such a way as to retain all natural regulatory elements of the gene (42). Furthermore, these constructs generate a fusion protein between the enzyme and the IEG protein that still translocates to the nucleus. Thus, this fusion gene drives expression of nuclear b-galactosidase, which can be readily detected in frozen sections by routine histochemistry. The nuclear localization was chosen, because it provides a more useful anatomical marker that discriminates pre- from postsynaptic elements. In addition, it concentrates the signal, thereby improving the signal-to-noise ratio.

We have used such fusion genes to construct cell lines (42) and transgenic mice (45) that accurately recapitulate expression of several IEGs (28, 44, 45, 46). This method provides excellent cellular resolution; it is relatively rapid compared to both immunohistochemistry and in situ hybridization; it can be made quantitative; it does not require specialized (and often expensive) reagents; it is nonradioisotopic; and there is no ambiguity with regard to the gene that has been activated. Using transgenic mice, one can obtain complete maps of fos僕acZ expression following administration of various types of stimuli to the central nervous system (45). Furthermore, primary neural cultures can be generated from the mice that permit analysis of gene expression in vitro following administration of classical neurotransmitters and neuropeptides, as well as neuropharmacological agonists and antagonists (45). In addition, permanent cell lines harboring the fusion gene can be grown in microtiter plates so that they can be treated, lysed, incubated with substrate, and assayed for b-galactosidase by a multiwell reader (42). This assay has a rapid throughput, it is quantitative, and it is particularly useful in screening for agents that interact with ion channels or neurotransmitter receptors.

The obvious caveat of this transgenic approach is that one must firmly establish that the transgene faithfully reflects the expression of the endogenous gene (although for some applications this may be irrelevant). To some degree this can be achieved by ensuring that the transgene is expressed under all of the same circumstances that have been reported in the literature for the cognate IEG. At least to a first approximation, this is true, and those differences that do exist are likely to reflect shortcomings in antibody specificity and/or species differences (most analyses of IEG expression have been performed on the rat). Thus both basal and inducible sites of Fos僕acZ expression are coincident with that observed using antibodies and oligonucleotides probes.

Another word of caution should be raised here regarding variability in expression between independent lines of mice harboring a particular gene construct. When embarking on this type of approach, one needs to be aware that rarely are two lines identical in their expression properties. Typically, there is great variability in the absolute level of expression. Thus for practical reasons we select lines that have robust expression. In addition, with fos僕acZ constructs we have noted that the induced patterns of expression are incomplete in some lines. That is, one observes induced expression but only in a subset of cells or tissues. Therefore, we apply a second selection criteria: namely that the overall pattern of expression should match, as far as we can determine it, the normal expression profile of the endogenous gene. For this reason, we always attempt to confirm novel sites of Fos僕acZ expression by some independent means such as immunohistochemistry or Northern blot (28). Based on our experience with over 20 lines of fos僕acZ rodents, one needs to derive a minimum of five independent founder lines to obtain one that fulfills both selection criteria of robust and faithful expression.

The half-life of the Fos僕acZ fusion protein is somewhat longer than that of Fos (42), although it is still relatively short-lived. The greater half-life of the fusion protein lengthens the time window in which expression can be detected (which is often an advantage for Fos), and it also improves the detection limits of the method. Despite the altered protein half-life, the transgene is induced and repressed with the same kinetics as fos. Like c-fos, the mRNA for the transgene is short-lived, a property that is conferred by inclusion of 3 untranslated sequences from c-fos in the fusion construct. A minor technical caveat is that the histochemical method can detect endogenous galactosidases, which makes working with tissues that are rich in these enzymes more difficult. Even though the brain is not among these tissues, we have, nevertheless, refined the fixation and incubation conditions to such a point that most tissues can now be studied. We should also point out that since the Fos僕acZ carries a nuclear localization signal, it is readily distinguished from artifactual staining which is invariably extra- or perinuclear.

A number of novel features of IEG expression have emerged from studies of these transgenic mice. Besides the well-documented induction of fos and jun by various stimuli, we found many tissues that either constitutively expressed these genes in the adult or that spontaneously expressed them during development (28, 44, 45, 46). In some instances the continuous expression of Fos僕acZ was associated with cells that were undergoing terminal differentiation, particularly where this culminated in cell death within a period of a few days to a week (45, 46). Examples of this in the adult animal included the skin and hair follicle, hypertrophic chondrocytes of the bone and follicular cells within atretic follicles of the ovary. Similarly a number of cell populations that are eliminated during development spontaneously expressed Fos僕acZ, such as interdigital web cells and cells of the periderm and the heart valve cushion. In other instances of fos僕acZ expression during development, the cells appear to undergo complex processes that include programmed cell death, migration, or transdifferentiation. Examples here include, the medial edge epithelium of the palate, the ureteric bud of the kidney, and developing bronchioles of the lung (28, 46). A number of examples of Fos僕acZ expression were also found in the developing nervous system (44). For instance, Fos僕acZ was noted in some of the cranial and dorsal root ganglia at a time when programmed cell death is known to occur. Other instances of naturally occuring Fos僕acZ expression appeared to be related to maturation of certain neuronal pathways and may be a reflection of plasticity. Thus in the perinatal brain, spontaneous Fos僕acZ expression is associated with cells in the olfactory tract, limbic system, and thalamus (44). This expression is more consistent with the acquisition of behaviors involving chemosensory, exploratory, and motor processing. It is now interesting to examine whether the transgene can be used to map the maturation of functional circuitry in the nervous system.

The circumstantial association of Fos僕acZ expression with programmed cell death, both in the nervous system and elsewhere, suggested that the transgene might have utility in following the demise of neurons under pathological conditions. This notion has been tested in three models of neuronal death. The first involved crossing the fos僕acZ transgenic mice onto the weaver mutation, which exhibits degeneration of cerebellar granule cells in the immediate postpartum period and neurons of the substantia nigra in the adult. In both cases, inappropriate expression of Fos僕acZ was observed in the affected neuronal populations (46).

In a second paradigm, the sciatic nerve was transected in the neonatal fos僕acZ mouse. This results in the degeneration of sensory neurons in the dorsal root ganglia and a-motor neurons in the ventral spinal cord. Once again, both populations of cells exhibited inappropriate expression of Fos僕acZ, suggesting that death triggered by growth factor deprivation results in an activation of the IEG cascade (46).

Finally, transgenic mice were exposed to the excitatory neurotoxin, kainic acid, which resulted in the delayed death of pyramidal neurons in CA1 and CA3 of the hippocampus as well as neurons within the amygdala. Again, all of these cells expressed Fos僕acZ while the death process was going on. Indeed, in this case, there were two phases of transgenic expression. There was an initial, transient, global phase of expression associated with the seizures elicited by kainic acid and a subsequent reexpression of Fos僕acZ some 4 to 7 days later only in the vulnerable neuronal populations. Furthermore, a hallmark of this delayed expression was that the LacZ was detected both in the nucleus and cytoplasm of the neurons (46). This may be because the cells were deteriorating and loosing nuclear integrity or simply were metabolically compromised and could not traffic the protein adequately. Alternatively, it may be that elevated proteolysis associated with the treatment results in the cleavage of some proportion of the LacZ from Fos, thereby removing the nuclear localization sequences. Whatever the reason, the presence of cytoplasmic Fos seems to be a harbinger of neuronal death. The key issue now is to determine what role, if any, the IEGs play in promoting or combating neuronal death in vivo.

The transgenic approach has one unique advantage over all other conventional neuroanatomical mapping techniques: namely, it can provide information regarding the signaling pathways operating in given neurobiological responses in vivo. The promoter of c-fos (and many other IEGs) contains a series of response elements that confer inducibility by various intracellular signaling pathways (reviewed in refs. 31 and 43). For example, experiments involving transient transfection in cell culture have established that nerve growth factor and NMDA induction of c-fos requires an intact serum response element (SRE). In contrast, depolarizing stimuli, or agents that act by elevating intracellular cAMP levels, act through a quite distinct site, the calcium-cAMP睦esponse element (CARE). This situation affords the possibility of dissecting these pathways in vivo by introducing point mutations into one or another of the responsive elements in the context of fos僕acZ transgenes.

Transgenic mice have been generated that carry a number of mutated fos promoter constructs driving lacZ expression. These mice reveal a number of properties of fos expression that provide insights into the cellular and biochemical mechanisms involved in neurophysiological and neuropharmacological responses. For example, some mutations eliminate both basal and stimulated expression of the transgene. In other cases, basal and continuous sites of expression are lost, but the gene is perfectly inducible in the brain. Yet other mutations reveal a heterogeneity among neuronal populations with regard to their response to a given stimulus. For example, with the unmutated promoter, all CA1 and CA3 pyramidal neurons of the hippocampus are induced for fos僕acZ by kainic acid, whereas one promoter mutation shows normal (if not better) induction in CA1 but no expression in CA3 following administration of the same stimulus. One would conclude, therefore, that the signaling pathways linking a single type of stimulant, kainic acid, to IEG expression is distinct in CA1 and CA3 neurons. Thus this genetic dissection approach offers a unique potential for furthering our understanding of information processing in specific neuronal populations.

We have now generated a series of transgenic mice and rats that harbor various fusion constructs between IEGs and lacZ. These animals provide a unique and novel resource for studying IEG expression in vivo. The approach can be extended by crossing the mice onto various neurologically mutant backgrounds or by using promoter mutations that can reveal subtleties in neuronal signaling that could not be inferred from any other known method. We feel that these animals represent merely the first generation of such constructs and that we have not yet begun to tap the real potential of this type of methodology. For example, parallel strategies might be to use the IEG promoters to drive inducible expression of engineered forms of IEGs that interfere with the function of the cognate proteins. Such transdominant suppressors might thereby provide a means by which one could establish whether individual IEG products contributed to particular neurobiological processes. Indeed, it now seems most likely that the emphasis in the field will shift from simple mapping studies to those involving function.

J. I. Morgan and T. E. Curran: Roche Institute of Molecular Biology, Roche Research Center, Nutley, New Jersey 07110.

published 2000