The LIM domain is a zinc-binding amino acid motif that characterizes various proteins which function in protein–protein interactions and transcriptional regulation. Expression patterns of several LIM protein genes are compatible with roles in vertebrate CNS development, but little is known about the expression, regulation, or function of LIM proteins in the mature CNS. Lmo1, Lmo2,and Lmo3 are LIM-only genes originally identified as putative oncogenes that have been implicated in the control of cell differentiation and are active during CNS development. Using in situ hybridization for mRNA and immunohistochemical detection of reporter protein expression in transgenic mice, we found thatLmo1, Lmo2, and Lmo3 show individually unique but partially overlapping patterns of expression in several regions of the adult mouse forebrain, including hippocampus, caudate putamen, medial habenula, thalamus, amygdala, olfactory bulb, hypothalamus, and cerebral cortex. In the hippocampal formation,Lmo1, Lmo2, and Lmo3 show different combinatorial patterns of expression levels in CA pyramidal and dentate granule neurons, and Lmo1 is present in topographically restricted subpopulations of astrocytes. Kainic acid-induced limbic seizures differentially regulated Lmo1, Lmo2, andLmo3 mRNA levels in hippocampal pyramidal and granule neurons, such that Lmo1 mRNA increased, whereasLmo2 and Lmo3 mRNAs decreased significantly, with maximal changes at 6 hr after seizure onset and return to baseline by 24 hr. These findings show that Lmo1, Lmo2, and Lmo3 continue to be expressed in the adult mammalian CNS in a cell type-specific manner, are differentially regulated by neuronal activity, and may thus be involved in cell phenotype-specific regulatory functions.
- LIM-only proteins
- transcriptional regulation
- limbic seizures
- gene expression
- cell phenotype
LIM proteins are a diverse group of transcription factors, proto-oncogene products, cytoskeletal proteins, and constituents of adhesion plaques that contain one or more cysteine-rich amino acid sequence motifs known as LIM domains (Way and Chalfie, 1988;Freyd et al., 1990; Karlsson et al., 1990; Sanchez-Garcia and Rabbitts, 1994). More than 40 different eukaryote LIM proteins have been identified in various tissues across a wide range of species, and the LIM domain sequence is highly conserved among these (Sanchez-Garcia and Rabbitts, 1994; Dawid et al., 1995). The LIM domain binds zinc ions and shows structural similarities to DNA-binding zinc finger proteins (Michelsen et al., 1993; Archer et al., 1994); however, most evidence favors a role for the LIM domain in protein–protein interactions (Rabbitts and Boehm, 1990; Schmeichel and Beckerle, 1994).
Several classes of LIM proteins have been identified in which one or more LIM motifs are present either alone (LIM-only proteins) or in association with recognized protein regions that may confer specific functions, such as DNA-binding homeodomain regions, cytoskeletal proteins, protein kinase domains, and regions that bind other proteins (Freyd et al., 1990; Crawford et al., 1994; Sanchez-Garcia and Rabbitts, 1994; Dawid et al., 1995; Taira et al., 1995; Weiskirchen et al., 1995). The functions of LIM-only proteins are less well understood. Homology between LIM domains suggests that they may be involved in LIM–LIM interactions (Rabbitts and Boehm, 1990) through which they may indirectly affect transcription (Sanchez-Garcia et al., 1993), as suggested by their putative roles in the regulation of cell differentiation or tumorigenesis (Boehm et al., 1991a; Warren et al., 1994).
Lmo1, Lmo2, and Lmo3 comprise a family of genes encoding LIM-only proteins (Foroni et al., 1992; Sanchez-Garcia and Rabbitts, 1994). Lmo1 and Lmo2 (originallyrbtn1, -2 or tgt1, -2) were identified as highly expressed transcription units near chromosomal translocation breakpoints associated with T-cell tumors (Boehm et al., 1988; McGuire et al., 1989; Boehm et al., 1991a). The LMO1 and LMO2 proteins have 50% amino acid identity, and LMO3 was identified subsequently on the basis of sequence homology (Foroni et al., 1992). Lmo1, Lmo2, and Lmo3 are all highly expressed in the developing mouse CNS (Boehm et al., 1991b; Foroni et al., 1992).Lmo1 has an unusual structure, with two promoter regions designated 1 and 1a (Boehm et al., 1990, 1991b). Transgenic mice expressing the bacterial lacZ gene fromLmo1-promoter 1 show β-galactosidase (β-gal) activity in cells in various regions of the developing CNS (Greenberg et al., 1990).
Considerable evidence now indicates that a number of LIM protein genes, including LIM-only genes, are active during development of the vertebrate CNS (Foroni et al., 1992; Tsuchida et al., 1994; Lumsden, 1995; Shawlot and Behringer, 1995). Less is known about the extent of expression, regulation, or function of LIM proteins in the mature vertebrate CNS. In this study we have compared the regional distribution and cellular patterns of Lmo1, Lmo2, andLmo3 expression in the hippocampus and other forebrain regions of adult mice by in situ hybridization using radiolabeled antisense oligonucleotide probes. To investigate whether expression of these LIM protein genes might be regulated by changes in neuronal activity, we examined the effects of kainic acid (KA)-induced limbic seizures (Ben-Ari, 1985) on relative levels of their mRNAs in the hippocampus, a site of considerable plasticity of gene expression in response to seizures in adult rodents (Bendotti et al., 1994; Morgan et al., 1987; Gall and Isackson, 1989). In addition, the hippocampal distribution of Lmo1 mRNA was compared with that of staining for β-gal in adult mice expressing the bacterial lacZ gene from the Lmo1-promoter 1 (Greenberg et al., 1990).
MATERIALS AND METHODS
Animals. Wild-type BKTO mice weighing 25–30 gm (Bantin-Kingman, Freemont, CA) were used for evaluation by in situ hybridization with and without KA-induced limbic seizures. The generation of transgenic mice expressing the bacteriallacZ gene from the Lmo1-promoter 1 has been described (Greenberg et al., 1990); these mice were used for histochemical evaluation with and without KA-induced limbic seizures. Animals were housed in small groups, maintained in a constant 12 hr light/dark cycle with controlled temperature and humidity, and allowed free access to food and water. All experiments were conducted in accordance with the United Kingdom Animal Scientific Procedures Act, 1986.
Drug administration. Animals were placed in the experimentation room 24 hr before drug injection to allow them to acclimatize. KA (30 mg/kg in 0.9% NaCl) (Sigma, St. Louis, MO) or vehicle (0.9% NaCl) were administered intraperitoneally. Animals were kept under observation for 1 hr and then returned to the home cage. Wild-type animals analyzed by in situ hybridization were killed by decapitation at set time intervals. The brains were rapidly frozen in n-pentane at −30°C and stored at −70°C before sectioning. Transgenic animals analyzed by histochemistry were killed by terminal barbiturate anesthesia followed by transcardiac perfusion with fixative, as described below.
In situ hybridization. Sections (10 μm, three sections/slide) were cut at −20°C using a cryostat (Bright Instruments) and thaw-mounted onto sterile poly-l-lysine-coated slides. Sections were air-dried and fixed using 4% paraformaldehyde in PBS for 5 min, rinsed twice in 1× PBS for 1 min, dehydrated in 70% ethanol for 5 min, and then stored in 95% ethanol at 4°C until use. Slides containing control and experimental sections were hybridized overnight in hybridization buffer, and [35S]dATP-labeled probes were diluted to a concentration of 3000 cpm/ml hybridization buffer. Sections were covered with parafilm and hybridized at 42°C in a humid atmosphere. Excess unbound probe was removed using the following stringency washes: 1× standard saline citrate buffer (SSC) at room temperature for 30 min and then 1× SSC at 55°C for 30 min. Slides were finally rinsed in 1× SSC, dehydrated in ethanol, and air-dried. Sections from control and experimental animals were apposed to the same sheet of Biomax MS film for a period of time before the film was developed under standard conditions. Nonspecific binding was determined using 100-fold excess of unlabeled probe. The length of time that sections were apposed to film was dependent on the mRNA under study, but once determined this time was kept constant for that particular probe. Care was taken to standardize procedures for control and experimental section at all stages. A selection of slides were coated with emulsion (LM1; Amersham, Arlington Heights, IL), and after 6 weeks in a light-proof box were developed, stained, and counterstained with hematoxylin and eosin for microscopic examination.
Northern blot analysis. Wild-type animals for Northern analysis were killed by decapitation, and RNA was extracted from mouse brain areas pooled from five animals as described (Sambrooke et al., 1989). Brains were homogenized (Polytron homogenizer) in guanidium thiocyanate buffer, and the total RNA was pelleted by centrifugation on a cesium chloride cushion at 32,000 rpm for 24 hr at 20°C (Beckman L8–70M ultracentrifuge with SW41Ti swinging-bucket rotor). After resuspension in Tris-EDTA buffer containing 0.1% SDS, the RNA was reprecipitated in 100% ethanol and sodium acetate at 4°C overnight and collected by centrifugation at 12,000 × g at 4°C for 10 min. The RNA was diluted to 20 μg/4.5 μl in diethyl pyrocarbonate-treated water, and then 20 μg of RNA was loaded per lane. This was separated using a denaturing 1% agarose/formaldehyde gel (Sambrooke et al., 1989) in 1× 3-[N-morpholino]propanesulfonic acid buffer along with standard RNA markers. Ethidium bromide was added to the gel to visualize the RNA after electrophoresis. The samples were electrophoresed at 60 V for 2.5 hr (Bio-Rad submarine gel electrophoresis tank; Bio-Rad, Hercules, CA), and the resulting gel was photographed under ultraviolet light against a fluorescent ruler (Polaroid Land camera and Polaroid type 57 film). The separated RNA was transferred overnight from the gel onto a nitrocellulose membrane by capillary elution, using 10× SSC buffer. The membrane was baked under vacuum for 2 hr at 80°C to fix the RNA to the membrane. After membranes were prehybridized in hybridization buffer overnight (Hybaid mini-hybridization oven), [32P]-labeled oligonucleotide probe was added at a concentration of 3 × 106 cpm/ml hybridization buffer and hybridized at 42°C overnight. Membranes were washed in 1× standard sodium chloride EDTA (SSPE)/0.1% SDS for 30 min at room temperature, followed by 1× SSPE/0.1% SDS for 30 min at 55°C. After they were wrapped in clingfilm, membranes were exposed to x-ray film (Kodak X-OMAT AR) at −70°C overnight. The size of the transcripts was calculated by reference to standard RNA markers.
Oligonucleotide probes. Oligonucleotide probes were made on an Applied Biosystems (Foster City, CA) synthesizer. After deprotection at 55°C for 16 hr, the probes were dried in a Speed Vac centrifuge, resuspended, purified, and dissolved in sterile water to give a stock solution of 20 ng/ml. They were labeled at the 3′ end using terminal deoxynucleotidyl transferase (New England Nuclear) and [35S]dATP (1000 Ci/mmol) or [32P]dATP (300Ci/mmol). The radiolabeled probe was purified using Sephadex columns (Biospin6), and dithiothreitol was added to a final concentration of 50 mm for [35S]dATP-labeled probes. Sequences of the antisense oligonucleotide probes used in this study were selected from nonoverlapping regions of the Lmo1, Lmo2, andLmo3 genes (Boehm et al., 1991a; Foroni et al., 1992). The following antisense oligonucleotide sequences (kindly provided by Dr. T. H. Rabbitts, Cambridge) were used: Lmo1, 5′-GCA GAC GGA CAG ATG GAC CTG GAG GCC AGA TGG TGG GCG TTA CTG-3′, complimentary to nucleotides 586–631; Lmo2, 5′-GAC TCT GGG CTA GAT GAT CCC ATT GAT CTT GGT CCA CTC GTA GAT-3′, complimentary to nucleotides 810–855; and Lmo3, 5′-ATG TAG TGC TTT GCA TTG TTG GGG AGA CGC TGC TGC CCC CTC ACC-3′, complimentary to nucleotides 705–750.
Analysis of autoradiographical data. Analysis of in situ hybridization autoradiograms was carried out using an MCID image analyzer (model M4) to measure relative optical density (ROD) values. Measurements were taken within the linear range of optical density levels, and in situ hybridization results are expressed in arbitrary units (ROD values ×100). Measurements were taken from at least five animals with three to six sections measured per animal, and the measurements from each animal were processed independently on the MCID. The final data were statistically analyzed using the nonparametric Kruskal–Wallis test for multiple measures, followed by Dunn’s post test for post hoc pair-wise comparisons (GraphPad Software). p < 0.05 was taken as the minimum level of statistical significance.
Histochemical analysis. Perfusion-fixed transgenic brains were allowed to sink in 25% buffered sucrose, and frozen sections were prepared at 40 μm. For routine immunohistochemical detection of β-gal, the fixative used was 4% paraformaldehyde. Free-floating tissue sections were incubated in a polyclonal rabbit antiserum against β-gal (Cappell, West Chester, PA) followed by the avidin–biotin–peroxidase method (Vectastain ABC kit; Vector Laboratories, Burlingame, CA). Immunoreactivity was visualized using diaminobenzidine (Sigma) as the chromagen. For double-labeling immunohistochemistry of β-gal (brown) and glial fibrillary acidic protein (GFAP) (blue), β-gal was stained first as just described, using diaminobenzidine as the brown chromogen. Sections were then washed briefly and stained using the same protocol with a primary rabbit antiserum against GFAP (Dako, Carpinteria, CA), and Vector-SG (Vector) as a blue chromogen. Timm staining of zinc-containing fibers was conducted as described (Geneser et al., 1993). For double Timm/immunohistochemical staining, the fixative used was that for Timm staining, which gave somewhat poorer but identifiable immunohistochemical staining for β-gal. After fixation, the sections were immunohistochemically stained first, followed by the Timm stain procedure. All stained sections were mounted on gelatin-subbed slides, air-dried, cleared in xylene, and coverslipped in DePe-X.
Oligonucleotide probe specificity
Lmo1, Lmo2, and Lmo3 share considerable sequence homology (Foroni et al., 1992), and the specificity of the in situ hybridization reaction obtained with the antisense oligonucleotide probes used in this study was tested and verified in several ways. The hybridization patterns obtained with each probe in the mouse forebrain were unique and included a number of nonoverlapping areas (Figs. 1, 2; Table1). In addition, the hybridization signal obtained with each radiolabeled probe was eliminated by competition with excess amounts of its corresponding unlabeled antisense oligonucleotide, whereas competition with excess amounts of the other two unlabeled oligonucleotides did not diminish the hybridization signal (Fig. 1). Northern blot analysis further showed that each of the probes bound to transcripts of different sizes, in agreement with previous studies that used probes derived directly from cDNA (Fig. 3).
Lmo1, Lmo2, and Lmo3 mRNA in mouse forebrain
Lmo1, Lmo2, and Lmo3 showed individually unique but partially overlapping patterns of expression in many regions of the adult mouse forebrain. Prominent sites of expression included various portions of the hippocampal formation, caudate putamen, medial habenula, thalamus, amygdala, olfactory bulb, hypothalamus, and cerebral cortex (Figs. 1, 2). Table 1 shows qualitative estimates of the relative intensity of the labeling obtained with individual Lmo mRNA probes in various forebrain regions, prepared by comparison of five untreated mice hybridized with all three probes in an alternate series of sections. The comparison is only between the expression levels of each individual gene relative to itself in different anatomical regions. It is not possible using in situ hybridization to compare the expression levels of different genes with one another, and no attempt has been made to do so. Thus, for example, the level of Lmo1expression is compared across various anatomical regions, but the expression levels of Lmo1, Lmo2, and Lmo3 within an anatomical region are not compared with one another.
The forebrain distribution patterns varied for the threeLmo genes such that Lmo3 showed the most widespread distribution, followed by Lmo1 and thenLmo2 (Figs. 1, 2; Table 1). The caudate putamen showed high relative levels of expression for both Lmo1 andLmo3, whereas Lmo2 was essentially undetectable, as determined by both in situ hybridization and Northern blot analysis (Figs. 1, 2, 3). Interestingly, Northern blot analysis showed approximately equal activities from Lmo1 promoters 1 and 1a in the caudate putamen (Fig. 3). In the amygdaloid complex,Lmo3 showed prominent expression in several nuclei, particularly the basolateral amygdala, whereas Lmo1 showed little and Lmo2 showed no expression (Figs. 1, 2; Table 1). In the thalamus, Lmo2 expression was prominent in the parafascicular nucleus, paracentral and other intralaminar nuclei;Lmo1 was prominent in the reticular and Lmo3 in the lateral geniculate nuclei (Figs. 1, 2; Table 1). In most of these areas, the size of the labeled cells in emulsion-dipped sections suggested that they were neuronal (Fig.2 D–F). Nevertheless, some small cells that may have been glial cells were also labeled. For this report, the expression patterns of Lmo1, Lmo2, and Lmo3 in the hippocampal formation and their changes after limbic seizures are described in greater detail.
Lmo1, Lmo2, and Lmo3 expression in hippocampus
All three Lmo genes showed prominent expression in different portions of the hippocampal formation. Hippocampal principal cells, i.e., pyramidal neurons of CA1, CA2, CA3 and subiculum, and granule neurons of the dentate gyrus, showed substantial but different combinatorial patterns of Lmo1, Lmo2, andLmo3 mRNA expression by in situ hybridization (Figs. 2, 4; Table 1). Hippocampal expression was confirmed for all three Lmo genes by Northern blot analysis (Fig. 3). In situ hybridization showed high relative levels of expression of Lmo1 in pyramidal neurons of CA2 and in dentate granule neurons, and moderate levels in pyramidal neurons of CA1, CA3, and subiculum (Figs. 2, 4; Table 1). Northern blot analysis of hippocampal tissue showed evidence of activities from bothLmo1 promoters 1 and 1a, with a slight predominance of activity from promoter 1 (Fig. 3). In situ hybridization showed about equal relative levels of expression of Lmo2 in pyramidal neurons of CA1, CA2, and subiculum, with somewhat less expression in dentate granule neurons and CA3 pyramidal neurons (Figs.2, 4; Table 1). In situ hybridization showed high relative levels of expression of Lmo3 in pyramidal cells in CA1, CA2, and the subiculum, as well as by dentate granule neurons, with somewhat less expression in CA3 pyramidal neurons (Figs. 2, 4; Table 1). Microscopic examination of sections dipped in photographic emulsion confirmed at the cellular level the observations made on autoradiographic film, and showed that within a particular subgroup of principal neurons (i.e., CA1, CA2, etc.) the individual expression levels of Lmo1, Lmo2, or Lmo3 were broadly similar across most neurons, so that differences between subgroups reflected essentially the whole group (Fig.4 D–F,H). Examination of emulsion-dipped sections also showed that Lmo1 and Lmo3, but notLmo2, were expressed by presumptive interneurons in various hippocampal regions (Fig. 4 G).
The distribution of Lmo1 mRNA was also compared with the distribution of immunoreactive β-gal in six untreated adult transgenic mice expressing lacZ fromLmo1-promoter 1. In these transgenic mice, most pyramidal neurons in the subiculum and CA1 and nearly all pyramidal neurons in CA2 were β-gal-positive (Fig. 5 A,D,E). These findings correlated well with observations of Lmo1mRNA as detected by in situ hybridization (Figs. 2, 4). In contrast, no pyramidal neurons in CA3 and no granule neurons in the dentate gyrus were β-gal-positive in transgenic mice (Fig.5 A,E,F), suggesting that Lmo1 promoter 1 was not active in these neurons and that the moderate expression levels of Lmo1 in CA3 pyramidal neurons and the more robust expression observed in dentate granule neurons by in situhybridization were attributable to activity of promoter 1a. This possibility is further strengthened by observations from Northern blot analysis indicating that although activity from Lmo1promoter 1 predominates somewhat, promoter 1a is also active in the adult hippocampus (Fig. 3). In the transgenic mice, scattered interneurons were also β-gal-positive throughout the hippocampus (Fig. 5 A,D,E). In addition, numerous cells that were stained positively for β-gal had the morphological appearance (Kosaka and Hama, 1986) of astrocytes in the molecular layer and hilus of the dentate gyrus or of radial astrocytes in the subgranular zone (Fig.5 A,F,I–L). β-Gal-stained presumptive astrocytes had a tufted appearance, with a small central cell body and many fine branches (Fig. 5 I). These cells, as well as radial glia, stained positively for both GFAP and β-gal in double-labeled sections (Fig. 5 J–L), confirming their identity as astrocytes. Some small β-gal-positive cells, which may have been progenitor cells that are known to exist in this region, were observed in the dentate subgranular zone (Suhonen et al., 1996); further analysis will be required to establish the identity of all β-gal-positive cell types. All cell types identified in the transgenic animals could be identified in emulsion-dipped sections processed for in situ hybridization detection ofLmo1, including pyramidal neurons in subiculum, CA1, and CA2, as well as presumptive interneurons and small potential glia cells (Fig. 4 D,G–I), strongly supporting the likelihood that β-gal expression in these transgenic mice reflects sites of endogenous expression of Lmo1 from promoter 1. In addition, previous analysis has shown that progeny of five different transgenic founder mice generated with this fusion gene construct all showed similar developmental patterns of β-gal staining, further indicating that the transgene expression patterns observed are not artifact (Greenberg et al., 1990). In comparing the pattern of labeling obtained by in situ hybridization with that of reporter gene expression in these transgenic mice, it should be noted that the immunohistochemical reaction product for β-gal was often deposited not only in the cell body but also throughout its processes, particularly in astrocytes. Thus, when viewed at low power as in Figure5 A,F, the degree of reporter protein expression visualized considerably overestimates the number of expressing cells.
To confirm that the differences in the staining observed between pyramidal neurons in different CA regions genuinely reflect anatomically recognized borders, we combined immunohistochemical analysis of β-gal with Timm staining. Timm stain identifies projections of the mossy fibers that pass from the dentate granule cells to CA3 cells and provides an accurate and well characterized means of identifying the CA2–CA3 border (Haug et al., 1971; Amaral and Dent, 1981; Swanson et al., 1987). Combined detection of β-gal and Timm in neighboring (Fig. 5 A,B) and the same sections (Fig.5 C) clearly demonstrated that Lmo1 expression from promoter 1 observes the CA2–CA3 border and is high in CA2 and absent from CA3 pyramidal neurons.
Regulation of Lmo1, Lmo2, and Lmo3 mRNA by seizure activity
To investigate whether hippocampal Lmo1, Lmo2, and Lmo3 expression levels might be regulated by changes in neuronal activity, we examined the effects of limbic seizure activity induced by intraperitoneal injections of KA. KA injections were dosed so that they resulted in behavioral activity such as freezing and rearing (Lothman and Collins, 1981; Ben-Ari, 1985) but did not cause generalized seizures and did not lead to hippocampal neuronal degeneration visible in tissue sections stained for cresyl violet after 7 d survival times. These KA-induced limbic seizures caused differential and statistically significant changes, as compared with vehicle-injected control animals, in the relative levels of Lmo1, Lmo2, and Lmo3 mRNA in hippocampal pyramidal and granule neurons as determined by densitometric analysis of sections processed by in situ hybridization (Figs. 6,7). Lmo1 mRNA increased and Lmo2and Lmo3 mRNAs decreased in both of these cell types in various hippocampal regions over a time course of 3–24 hr (Figs. 6,7). Few statistically significant changes were seen at 1 hr, and all changes were maximal at 6 hr after onset of seizure activity and had returned approximately to baseline by 24 hr. The changes can be summarized as follows. Lmo1 showed a very pronounced increase in expression in CA1 and particularly in CA2 pyramidal neurons, with a substantial increase in dentate granule neurons and little change in CA3 (Figs. 6 A, 7 A).Lmo2 showed a pronounced drop in expression to very low relative levels in CA1–CA3 pyramidal neurons and to essentially undetectable levels in dentate granule neurons (Figs.6 B, 7 B). Lmo3 showed a substantial decrease in expression but was still easily detectable in CA1, CA2, and CA3 pyramidal neurons and dentate gyrus granule neurons (Figs. 6 C, 7 C). Microscopic examination of sections dipped in photographic emulsion confirmed observations made on autoradiographic film and showed that within a particular subgroup of principal neurons (i.e., CA1, CA2, etc.), the individual expression levels of Lmo1, Lmo2, or Lmo3 were similar across most neurons, and that changes induced by limbic seizure activity in different subgroups reflected essentially the whole group. No changes in relative levels of Lmo1, Lmo2, and Lmo3 mRNA were observed qualitatively or quantitatively in the habenula, striatum, or cerebral cortex over this same time period (Figs. 6, 7). Cortical measurements were taken from the entire cortex overlying the hippocampus and may not have detected changes in particular layers or subregions. Although limbic seizure activity modulated expression levels of all three of these genes, seizure activity did not alter the basic pattern of cells expressing these genes; in particular, seizure activity did not induce Lmo gene expression in novel groups of CNS cells not already observed in untreated mice. In addition, four transgenic mice expressing lacZ from Lmo1promoter 1 were examined at 12 hr (to allow time for translation of protein) after KA-induced limbic seizures. The pattern of positive cell types in these animals was identical to that seen in untreated transgenic mice, and no novel sites of β-gal staining were observed.
Lmo1, Lmo2, and Lmo3 are a family of LIM-only genes originally identified as putative oncogenes (Boehm et al., 1991a) that are active during mammalian CNS development (Greenberg et al., 1990; Boehm et al., 1991b; Foroni et al., 1992). In this study we have shown that Lmo1, Lmo2, and Lmo3 continue to be expressed in a number of forebrain regions in the adult mouse, particularly in the hippocampal formation where they show different combinatorial patterns of expression in pyramidal neurons of CA1, CA2, CA3, and dentate granule neurons, as well as in restricted subpopulations of astrocytes. We also found that expression levels of these LIM-only genes in hippocampal principal neurons were differentially regulated in response to limbic seizure activity. Together these observations suggest that Lmo1, Lmo2, andLmo3 may be involved in cell phenotype-specific regulatory functions in the adult CNS.
LIM proteins and the adult CNS
LIM proteins represent a broad family of molecules whose functions in the CNS are not well understood. With their potential for protein–protein interactions and associations with various different functional domains such as DNA binding regions, cytoskeletal regions, or kinase domains, LIM proteins are candidates to be important regulatory molecules. A number of LIM proteins, including LIM-only proteins, seem to be involved in the control of cell differentiation in neural as well as non-neural tissues and in vertebrates and invertebrates (Way and Chalfie, 1988; Freyd et al., 1990;Sanchez-Garcia and Rabbitts, 1994; Warren et al., 1994). Such observations are consistent with findings that various LIM protein genes are expressed during, and may be involved in controlling, particular aspects of CNS development. In a well documented example, the combinatorial expression patterns of a family of LIM homeobox genes have been shown to define distinct subclasses of motor neurons in the spinal cord, and it has been suggested that these expression patterns play a prominent role in determining the topographic organization of motor projections in the developing vertebrate CNS (Tsuchida et al., 1994; Lumsden, 1995).
Less is known about LIM proteins in the adult CNS. The LIM-homeodomain protein ISL-1 is present in various regions of the adult rat brain (Thor et al., 1991), related Isl genes have been identified in mature fish brain (Gong et al., 1995), andKiz-1, which encodes a LIM-kinase protein, has been identified in mature mouse brain (Bernard et al., 1994). Factors affecting the regulation of these genes have not been reported, and the functions of these or other LIM proteins in the mature CNS are not known.
In this study we found prominent expression of the LIM-only genesLmo1, Lmo2, and Lmo3 in numerous regions of the adult mouse forebrain, including the hippocampal formation, caudate putamen, medial habenula, thalamus, amygdala, olfactory bulb, hypothalamus, and cerebral cortex. The results were confirmed with various specificity tests and were consistent across the three different types of techniques used: in situ hybridization, Northern blot analysis, and evaluation of transgenic mice. In most areas Lmo gene expression seemed to be neuronal. In the transgenic mice, some areas contained Lmo1-β-gal-positive cells that had the morphological appearance of astrocytes and were double-labeled with GFAP.
The role of LIM-only proteins in the adult CNS is not yet clear. LIM-only proteins have been proposed to function by protein–protein interactions, possibly also with other LIM proteins (Rabbitts and Boehm, 1990; Sanchez-Garcia et al., 1993; Sanchez-Garcia and Rabbitts, 1994). If this proves to be the case, then the widespread expression of LIM-only genes in the adult CNS suggests that other LIM protein genes may be active in the adult CNS as well. Additional studies will be required to determine this. Previous studies have shown thatLmo1, Lmo2, and Lmo3 are expressed in specific and restricted populations of cells in the developing mouse CNS, in the hindbrain, and in cerebellum and various forebrain regions, including olfactory bulb, caudate putamen, hippocampus, thalamus, hypothalamus, and neocortex (Greenberg et al., 1990; Boehm et al., 1991b; Foroni et al., 1992), in many of the same areas in which expression was demonstrated in adult mice in the present study. Although detailed comparisons have not yet been made, preliminary observations suggest that Lmo1, Lmo2, and Lmo3 are expressed in the same cell types in some brain regions, such as the hippocampus, from early development through to and persisting in the adult. Thus, these LIM-only genes may be involved not only in the development but also in the subsequent maintenance in the adult of specific cell phenotypes in the CNS. The combinatorial pattern of LIM-only gene expression in hippocampal neurons favors this possibility.
Lmo1, Lmo2, and Lmo3 and the phenotype of adult hippocampal cells
The function of the hippocampal formation depends on a precisely defined topographic arrangement of groups of neurons that exhibit well documented differences not only in cellular structure but also in afferent and efferent projections, intrinsic chemistry (e.g., of transmitters, growth factors, and receptors, etc.), and electrophysiological characteristics (Swanson et al., 1987; Amaral and Witter, 1989; Gall et al., 1991). In this study we found that the different combinatorial patterns of relative expression levels ofLmo1, Lmo2, and Lmo3 defined the well recognized subgroups of hippocampal principal neurons (Fig. 8). Pyramidal neurons of CA2 showed prominent expression of all three LIM-only genes, whereas CA1 pyramidal neurons showed somewhat lower relative expression levels. Pyramidal neurons of CA3 showed considerably lower relative expression levels of Lmo2, Lmo3, and Lmo1 from promoter 1a and no expression ofLmo1 from promoter 1. Granule neurons of the dentate gyrus show prominent expression of Lmo3 and Lmo1 from promoter 1a, but little expression of Lmo2 and no expression of Lmo1 from promoter 1 (Fig. 8). Importantly, major anatomical and functional boundaries that exist between well recognized groups of hippocampal pyramidal neurons, such as that between CA2 and CA3, were also precisely, and in some cases dramatically, defined by differences in relative expression levels of LIM-only genes. In addition, expression of Lmo1 from promoter 1 defined several topographically restricted subpopulations of astrocytes in the dentate gyrus. These findings suggest that LIM-only genes may be involved in defining subgroups of hippocampal cells. Interestingly, most of theLmo-expressing groups of principal neurons in the hippocampus did not seem to be identified by the expression of a singleLmo gene, but by different combinations of relative levels of Lmo gene expression. These findings are consistent with suggestions that combinatorial genetic events underlie the diversity of cell phenotype in the CNS (He and Rosenfeld, 1991; Struhl, 1991), and with reports that combinatorial expression patterns of a family of LIM homeobox genes distinguishes subclasses of developing motor neurons (Tsuchida et al., 1994).
Regulation of gene expression in the adult hippocampus after limbic seizure activity
Changes in transcription may be fundamental to certain hippocampal functions, such as its role in the formation of long-term memory (Squire, 1992; Bourtchuladze et al., 1994; Huang et al., 1994;Stevens, 1994; Sossin, 1996). In agreement with this possibility, hippocampal neurons show considerable plasticity of gene expression in response to changes in neural activity (Gall et al., 1991; Berzaghi et al., 1993; Bendotti et al., 1994), and factors produced by hippocampal neurons affect the phenotype and function of afferent neurons (Sofroniew et al., 1990; Thoenen, 1995). Activity-dependent changes in hippocampal gene expression induced by limbic seizures have been well characterized. After seizure activity, expression levels of immediate early genes in hippocampal principal neurons increase rapidly, peak at 1–2 hr, and decline rapidly (Morgan et al., 1987; Gass et al., 1993). Over a more prolonged time course that varies from 3–6 hr to 1–3 d and in some cases longer, expression levels of numerous other genes change, including genes encoding neuropeptides, neurotrophins, and growth associated proteins (Gall, 1988; Gall and Isackson, 1989; Gall et al., 1991; Bendotti et al., 1994; Lauterborn et al., 1995). All of these genes are expressed in a cell phenotype-specific manner, show differential levels of expression in different subsets of hippocampal principal neurons, and show changes in levels of expression in some but not other neuronal subgroups after seizure activity. In some cases expression levels increase after seizures, whereas in others they decrease.
In this study we found that expression levels of the LIM-only genes Lmo1, Lmo2, and Lmo3 are regulated by neuronal activity in a cell-specific manner, such that relative levels of Lmo1 increased and relative levels of Lmo2 andLmo3 decreased in different subgroups of hippocampal principal neurons after limbic seizure activity (Fig. 8). It is important to note that although changes in neural activity modulated expression levels of these genes, neural activity did not alter the basic pattern of cell type expressing these genes, such that the changes seen after seizure activity were specific to neurons already showing Lmo gene expression. Thus, combinatorial patterns ofLmo gene expression not only defined specific subsets of hippocampal neurons in untreated animals, but seizure-related increases in neural activity differentially regulated the relative levels of expression of these LIM proteins in different subsets of hippocampal neurons.
The functional role of the activity-dependent changes in expression levels of LIM-only genes that we observed in hippocampal principal neurons is not certain. As described above, several lines of evidence suggest that LIM-only proteins may have intracellular regulatory functions, perhaps by interacting with other LIM proteins that have functional domains such as DNA binding regions, kinase regions, or cytoskeletal regions. Taken together these observations suggest that the combinatorial expression patterns of Lmo1, Lmo2, and Lmo3 may be involved in the specification of defining characteristics to hippocampal neurons of particular phenotypes in the adult, and may differentially influence transcription or other regulatory processes in response to changes in neuronal activity in a cell phenotype-specific manner.
This work was supported by Parke-Davis, The Wellcome Trust, and Medical Research Council. We thank Dr. T. H. Rabbitts for oligonucleotide sequences, J. Drew for technical assistance, and J. A. Bashford, I. Bolton, and A. P. Newman for photography.
Correspondence should be addressed to M. V. Sofroniew, Medical Research Council Cambridge Centre for Brain Repair, Forvie Site, Robinson Way, Cambridge CB2 2PY, UK.