- Top of page
- Materials and methods
The limbic system, and in particular the amygdala, have been implicated in autism. The amygdala is a complex structure that in rodents consists of at least 12 different nuclei or subnuclei. A comparative analysis of amygdala neuroanatomy in normal vs. autistic brains would be aided by the availability of molecular markers to unambiguously recognize these different amygdala substructures. Here we report on the development of methods to identify genes enriched in the central, lateral and medial nuclei of the rodent amygdala. Our results suggest that laser-capture microdissection of specific amygdala subnuclei, when combined with linear amplification of cRNA probes for oligonucleotide microarray hybridization, can efficiently identify genes whose expression is confined to these substructures. Importantly, many of these genes were missed in previous gene expression-profiling experiments using whole amygdala tissue. The isolation of human orthologs of these subnucleus-specific genes, and/or the application of these methods directly to human tissue, may provide useful markers for characterizing neuropathological correlates of autism, as well as for identifying molecular differences between normal and autistic brains.
Over the last decade, much progress has been made in understanding the neurobiology of affective disorders. Animal models of several dysfunctions, including stress, depression and schizophrenia, have shed light on their etiology and modulating factors. In addition, the identification of many genes that play a role in the generation of or susceptibility to such disorders has deepened our understanding of the genetic players involved. More recently, the direct manipulation of some of these genes in transgenic and knockout mice has enabled both the establishment of new animal models (e.g., for polyglutamine repeat disorders (Burright et al. 1995; Mangiarini et al. 1996) and schizophrenia (Mohn et al. 1999)), as well as an examination of their influence on specific behaviors, such as in the modulation of fear responses by GABA (Crestani et al. 1999) or 5-HT (Gross et al. 2002) receptors.
To characterize a neurological disease at a molecular level, it is necessary to first identify the brain region that is affected by it. From the molecular biology perspective, it is therefore important to first pinpoint the genes that are normally expressed in these brain regions. Genes that are specifically expressed may play a direct and crucial role in the functioning of these areas. Alternatively, if their function is dispensable or their role unclear, the identification of such region-specific genes is nevertheless of utmost importance, because they may provide tools to investigate brain function and development in humans and in animal models of such diseases.
With these ideas in mind, we set out to find molecular markers for the amygdala, in order to have better tools to dissect its development and function. The amygdala is thought to control emotional behaviors, such as fear, anxiety and emotional learning (reviewed in Davis 1992; LeDoux 1995; Rogan & LeDoux 1996). A variety of lesion studies done in rodents or monkeys indicate that the amygdala is necessary for the evaluation of fearful stimuli. In addition, human patients with damage in the amygdala fail to recognize fearful visual or auditory stimuli (Adolphs et al. 1994; Scott et al. 1997), and also show general hypoemotionality. Since it is believed that some psychiatric conditions such as generalized anxiety or post-traumatic stress disorders are maladaptations of normal fear responses (Rosen & Schulkin 1998), the role of the amygdala in fear conditioning has attracted special attention (reviewed in Flint 1997; LeDoux 1995).
Amygdala dysfunction has been implicated in autism (Adolphs et al. 2001; Baron-Cohen et al. 2000; Sparks et al. 2002). Autism is defined behaviorally by social withdrawal, impaired verbal and non-verbal communication and restricted and stereotyped patterns of behavior and interests. Moreover, autistic patients typically have trouble inferring what other people think or feel and seem to live in a child-like, more simplistic emotional world. Similarly, patients with amygdala damage show some aspects of social impairment, such as general hypoemotionality and difficulty in recognizing fearful stimuli. Because of this similarity, it has been postulated that an impaired amygdala can partially account for some autistic symptoms (Adolphs et al. 1994; Sweeten et al. 2002). This hypothesis was supported both by postmortem analysis of autistic brains, which indicated increased neuronal packing in the cortical, central, medial and basal subnuclei of the amygdala (Kemper & Bauman 1993), and by brain imaging studies evidencing abnormal amygdala volume (although seemingly opposite findings have been observed, showing larger (Abell et al. 1999), and lower (Pierce et al. 2001) volumes). Furthermore, functional imaging studies have indicated reduced amygdala activation in autistic patients while processing photographs of a human face (Baron-Cohen et al. 1999; Pierce et al. 2001). When asked to describe the mental state of the photographed person, adults with high-functioning autism have deficits in this task, suggesting a functional amygdalar anomaly (Baron-Cohen et al. 1999). Finally, the only existing animal model of autism involves ablation of the amygdala. Amygdala lesions in rhesus monkeys produce aspects of autistic symptoms (reviewed in Baron-Cohen et al. 2000), while in neonatal rats they affect social and non-social behavior (Wolterink et al. 2001). The amygdala theory of autism therefore has many proponents. Nevertheless, autism is a complex disorder that probably involves malfunction of as many as 20 genes, which may interact with unknown environmental triggers, so it is not surprising that as yet not all autistic cases seem to bear the same pathology (Bailey et al. 1998).
Anatomically, the amygdala is a complex forebrain structure composed of over a dozen subnuclei (Pitkänen et al. 1997), such as the central, lateral, basomedial and medial subnuclei. These subnuclei have different functions. In general, there is a unidirectional informational flow, from lateral structures (such as the lateral and basolateral nuclei), to more medial ones (including the medial, basomedial, and central nuclei) (Pitkänen et al. 1997; Swanson & Petrovich 1998). The central and lateral nuclei have been suggested to play an important role in fear conditioning. The lateral nucleus is the site of sensory input convergence (Maren & Fanselow 1996; Pitkänen et al. 1997), while the central nucleus is the output. The central nucleus projects to hypothalamic and brainstem regions involved in the reactions to fear (Davis 1992), that produce perspiration, increased blood pressure and bradycardia. Finally, the amygdala possesses extensive connections with the neocortex, cholinergic basal forebrain, striatum and hippocampal formation (Amaral et al. 1992). Therefore it can also influence higher cognitive processes and perception.
As mentioned above, dysfunction of the amygdala has been implicated in autism. However, the neuroanatomical analysis of autistic brains (Bauman & Kemper 1988; Bauman & Kemper 1994; Courchesne 1997) is limited by our ability to identify specific subnuclei in the human amygdala, and to relate them to specific functions. Moreover, the resolution and contrast of most current magnetic resonance imaging (MRI) methodologies (Abell et al. 1999) is inadequate to discriminate between subtle amygdaloid subdivisions such as those of the medial nucleus. The availability of specific molecular markers for different amygdaloid subnuclei in humans would not only aid in the anatomical characterization of autistic amygdala from postmortem tissue samples, but might also provide tools to examine developmental abnormalities in these structures at earlier stages.
We previously described our initial efforts to characterize amygdala-enriched gene products using microarray technology (Zirlinger et al. 2001). There, we compared gene expression levels from five different relatively large and heterogeneous regions of the adult mouse brain, including the amygdala, cerebellum, hippocampus, olfactory bulb and periaqueductal gray. These first results were encouraging, as we were able to identify about 30 amygdala-enriched transcripts, and also confirmed their predicted expression patterns by in situ hybridization for a subset. In addition, we found that only 0.5% of the expressed genes were differentially expressed in each region. It is noteworthy that very similar figures were obtained by an independent study (Sandberg et al. 2000). This relatively high homogeneity among brain regions was somewhat unexpected for two reasons. First, neurons are very heterogeneous, with potentially hundreds of different neuronal cell types in the cerebral cortex alone (Serafini 1999). Second, early studies done prior to the advent of large-scale microarray technology had suggested that a large fraction of the genome was expressed in the brain (Milner & Sutcliffe 1983).
Two main reasons may explain the relatively low number of differentially enriched transcripts that we initially found. First, the use of high-density oligonucleotide arrays (Lipshutz et al. 1996) by definition limits the screen to genes and expressed sequence tags (ESTs) of previously known sequence. Therefore, if indeed highly specific brain transcripts exist, they may not have been cloned yet and thus represented on the arrays. Nevertheless, we decided that the convenience and proven reliability of commercial oligonucleotide microarrays merited their use, despite their relatively high cost. Second, highly specific, nonabundant transcripts may be lost due to simple dilution. Because of the cellular diversity in the brain, important expression differences occurring in a subpopulation of cells that make up a small fraction of the total population may simply be diluted out and undetected.
To try to solve the dilution problem, we decided to start with more finely dissected, homogeneous tissue, and search for genes differentially expressed within different amygdala subnuclei. Laser capture microdissection (LCM) has been developed to extract pure cell populations from specific regions of tissue sections, under direct microscopic visualization (Simone et al. 1998). A transfer film is applied to the surface of the tissue section placed on a standard glass slide, and is activated by a low-power laser beam. When activated, the film focally adheres to the cells of interest and thus permits their collection, while surrounding tissue that has not been submitted to the laser shot is left intact. In principle, the low energy laser used does not alter cellular contents, which can then be reliably collected for nucleic acid or protein extraction (Emmert-Buck et al. 1996).
We decided to conduct RNA-expression studies on laser-captured material from the lateral, central and medial subnuclei, because their roles are better defined than those of other amygdala nuclei and their anatomical demarcations are more clearly visualized with Nissl staining. We compared gene expression profiles not only between these subnuclei, but also compared those data with the profiles obtained in our previous screen, where a rather crude hand dissection of the whole amygdala was performed.
The results presented here show that such a combination of technologies, namely LCM-extraction of RNA from distinct amygdala subnuclei, followed by linear RNA amplification and microarray analysis, provides a useful tool to further characterize the molecular constituents of small neuronal subpopulations.
Materials and methods
- Top of page
- Materials and methods
We compared gene expression among three amygdala subnuclei: central, lateral and medial. The overall strategy consisted of the fine dissection of subregions by LCM with an Arcturus PixCell IIe instrument (LABTRADE Inc., Miami, FL), followed by two rounds of linear T7-based RNA amplification prior to probe synthesis. About 10–15 µg Biotinylated cRNA probe was then synthesized and hybridized to affymetrix U74v2 microarrays. These sets, comprising about 36 000 mouse genes and EST clusters (http://www.affymetrix.com/products/arraysspecificmgu74.affx), are composed of three subarrays (U74v2A, U74v2B, U74v2C). Three biotinylated cRNA probe replicates were synthesized for each subnucleus, from independent cDNA templates generated from a single RNA LCM-extraction from two one-month-old male C57BL/6J mice.
Microarray data were first normalized, and specific parameters calculated by the affymetrix software including average difference values, number of positive and negative cells (where ‘cell’ or ‘probe’ denotes one oligonucleotide within the set representing each gene on the microarray), and total number of probe pairs (perfect-match and mismatch) used to represent each gene on the microarray, were exported and analyzed with a custom software, written in matlab (the MathWorks, Natick, MA). The software was designed to identify genes that satisfy three criteria for enrichment in a given subnucleus:
an average difference value (ADV), a measure of transcript abundance, above a minimum threshold value
a large proportion of positive perfect-match cells, which indicates significant hybridization of the biotinylated cRNA probe to different oligonucleotides representing the gene
a fold-difference in expression compared to all replicates from the remaining subnuclei that is above a minimum threshold value.
Each of three replicates from each subnucleus was independently analyzed in this manner.
The threshold values for these enrichment parameters were determined empirically. We found that genes enriched in particular subnuclei can be typically identified with the following constraints:
a minimum ADV corresponding to ∼ 8% of the overall intensity of hybridization to the chip, a parameter used to normalize the datasets. In this case, the normalization parameter was set to 2,500, so the threshold ADV for an enriched gene was set to 200
a difference of (positive–negative) perfect-match probes greater than 6 or 3, depending on whether the number of probe pairs used to represent the gene was greater or smaller than 10, respectively
a ratio of expression compared to all other replicates from different subnuclei of at least 2.5.
These thresholds ensured both minimum detection level and sensitivity that could be measured and verified by an independent method, such as in situ hybridization. Detailed protocols for sample preparation, software description and data analysis, and in situ hybridization can be found in (Zirlinger in press).