Previous work in animal models has shown that projections from the basolateral amygdala (BLA) progressively infiltrate the medial prefrontal cortex (mPFC) from birth to adulthood, with the most dramatic sprouting occurring during the postweanling period. GABAergic (gamma-aminobutyric acidergic) interneurons in the human homolog of the rat mPFC have been implicated in the pathophysiology of schizophrenia, an illness with an onset that is delayed until late adolescence. Here we investigated the interaction of BLA fibers with mPFC GABAergic interneurons from postnatal day 6 (P6) to P120 using anterograde tracing and immunocytochemistry. We found a 3-fold increase in axosomatic and an 8-fold increase in axo-dendritic contacts in both layers II and V of the mPFC. Ultrastructural analysis using a colloidal gold immunolocalization demonstrated that the greatest proportion of BLA appositions were with GABA-negative spines (30.8%) and GABA-positive dendritic shafts (35.5%). Although GABA-negative interactions demonstrated well-defined axo-spinous synapses, membrane specializations could not be identified with confidence in GABA-positive elements. Our findings suggest that GABAergic interneurons are major targets for BLA fibers projecting to the mPFC. The establishment of this circuitry, largely during adolescence, may contribute to the integration of emotional responses with attentional and other cognitive processes mediated within this region during corticolimbic development.
The amygdala has been implicated in the mediation of memory, attentional responses, and emotional behaviors (Pitkanen et al. 1997); and its basolateral amygdala (BLA) in particular, play an important role in processing fear and anxiety (LeDoux 2000). The anterior cingulate cortex (ACCx, the human homologue of rat medial prefrontal cortex, mPFC) is believed to integrate emotional processes with cognitive function at the cortical level (Vogt et al. 1992; Allman et al. 2001); therefore, the development of the connectivity between the BLA and ACCx may be important in the etiology of mental illness. We previously reported that most BLA afferents in the mPFC of the rodent are established during the postweanling (or adolescent) period, corresponding to the developmental stage at which symptoms of many psychiatric diseases in humans manifest (Cunningham et al. 2002). Although this study found that BLA afferents predominantly target dendritic spines, presumably of pyramidal neurons, a large number of contacts were seen with dendritic shafts as well. Because most GABAergic (gamma-aminobutyric acidergic) interneurons are aspiny (Benes and Berretta 2001), we postulated that a substantial proportion of BLA afferents may be innervating GABAergic interneurons.
The GABAergic interneuron is the substrate for inhibitory tone within neural networks. GABAergic cells comprise up to 25% of the neurons in the brain (Tamamaki et al. 2003) and are integrated within virtually every neural circuit in the central nervous system. Modulation of large areas of cortex may be achieved through the direct activation of GABAergic interneurons via their extensive local connectivity. Indeed, an individual GABAergic interneuron can generate inhibitory synaptic potentials sufficient to simultaneously influence the activity of large numbers of striatal projection neurons (Koos and Tepper 1999). Serving as local modulators, GABAergic interneurons can receive afferents from diverse brain regions and may integrate this information to synchronize spontaneous and elicited activity and to regulate the final output of intrinsic neurons (Ramanathan et al. 2002).
The present study was designed to investigate the hypothesis that GABAergic interneurons may be a major target of BLA afferents, and may thus be positioned to modulate amygdalo-cortical excitatory activity. In the current report, BLA afferents were identified with an anterograde tracer, and their interaction with GABA-immunoreactive (-ir) elements within the mPFC was evaluated throughout postnatal development using single and double immunolabeling, together with brightfield, confocal, and electron microscopy (EM).
Subjects and Surgery
Litters of Sprague–Dawley rats were received at postnatal day 2 (P2) and were culled to 10 pups per dam. Pups were housed with their mother in clear plastic cages, handled daily prior to surgery, and weaned on P21. All animals were maintained on a 12-h light/dark schedule, with food and water provided ad libitum. Twenty-one male Sprague–Dawley rats were used for these studies. For confocal immunofluorescent evaluation, 7 developmental time points were examined: P6 (N = 2), P16 (N = 3), P25 (N = 2), P35 (N = 2), P45 (N = 3), P56 (N = 1), and P120 (N = 1). For bright field immunohistochemistry, 3 P90 animals were used and for EM, 4 P90 rats were used. Anesthesia and surgical procedures varied according to age of animals. Animals at P35 and younger were placed within a Cunningham neonatal adapter fitted to a Stoelting stereotaxic frame (Cunningham 1993, p. 61). Hypothermic anesthesia was utilized for P6 pups, achieved by covering animals with approximately 6 cm of crushed ice for a duration of 1 min/g body weight. Older animals were anesthetized with a combination of ketamine and xylazine and placed directly into a stereotaxic frame. Precise stereotaxic coordinates of the BLA were obtained by first creating an atlas for each time point (P45 and younger), then refining coordinates using stereotaxic microinjections of India ink. Biocytin (e-biotinoyl-L-lysine, Molecular Probes, Eugene, OR) 7.5% in 5 mM Tris buffer, pH 7.4, was pressure injected using a 1-μL Hamilton syringe fitted with a glass pipette pulled to a final tip diameter of 25–30 μm. Injection volumes varied according to age of animal, ranging from 100 to 600 nL, and were made at a rate of approximately 100 nL/min. Survival times ranged from 24 to 40 h, determined empirically, depending on stage of development and thus travel distance at each developmental time point. Following appropriate survival times, animals were deeply anesthetized and perfused transcardially with 0.9% saline solution, followed by 4% paraformaldehyde in 0.1 M phosphate buffer containing 0.5% gluteraldehyde. Perfusion solutions were maintained at 4 °C, pH 7.4, and perfused at rates corresponding to each subject's estimated cardiac output (5–50 cc/min). After removal from the skull, brains were postfixed in the same fixative for 12 h, and Vibratome sections were cut with a thickness of 50 μm in the coronal plane.
Free-floating sections were first rinsed in 0.1 M phosphate-buffered saline (PBS), pH 7.4, containing either 0.05% (EM) or 0.2% (bright-field/fluorescence microscopy) Triton X-100 (PBS-Tx). The tissue was blocked with 10% normal donkey serum in PBS-Tx/3% bovine serum albumin (BSA) for 60 min. Sections were incubated in rabbit anti-GABA antibody (1:1000, Sigma, St Louis, MO) 12 h at 4 °C. Sections were then washed 3 times with PBS-Tx for 10 min. This was followed by an avidin–biotin complex reaction against rabbit; GABA-positive cells were visualized using a diaminobenzadine/peroxidase reaction. Sections were then rinsed 3 times for 30 min in PBS, blocked for 1 h, and incubated in mouse anti-biotin antibody for (1:5000, Dako, Glostrup, Denmark) 12 h at 4 °C. Sections were washed 3 times with PBS-Tx for 30 min, exchanging buffer every 10 min. This was followed by incubation in biotinylated donkey anti-mouse antibody (1:500, Vector Laboratories, Burlingame, CA) for 1 h. Sections were rinsed again for 30 min, and placed in streptavidin–horseradish peroxidase complex (1:4000, Vector Laboratories, Burlingame, CA). Biocytin-containing fibers were visualized using a 3,3′-diaminobenzidine-nickel sulfate solution containing H2O2. Sections were washed thoroughly in PBS, mounted onto gelatin coated slides, air dried, dehydrated through a graded series of ethanol, cleared in xylene, and coverslipped with Permount medium. For double immunofluorescence, sections were incubated overnight with rabbit anti-GABA (1:5000, Sigma, St Louis, MO) combined with mouse anti-biotin (1:1000, Dako, Glostrup, Denmark). Fibers were then tagged with tetramethylrhodamine isothiocyanate-conjugated donkey anti-mouse (1:250, Vector Laboratories, Burlingame, CA) combined with fluorescein isothiocyanate-conjugated donkey anti-rabbit (1:250, Vector Laboratories, Burlingame, CA) to visualize GABAergic neurons.
For EM analysis, tissue was processed immunohistochemically as described above. Sections (40 μm thick) were postfixed with 1% OsO4 in sodium phosphate buffer, pH 7.2, dehydrated through a graded series of ethanol, and flat embedded in Embed-812 between ACLAR sheets. Appropriate areas were attached to epon blanks and sectioned at a thickness of 30–40 nm (silver-gray) using a Reichert–Jung ultramicrotome. Thin sections were picked up on nickel grids, air dried, and etched briefly with 1% sodium metaperiodate in PBS. The sections were washed in filtered double distilled water and incubated in 1% BSA in PBS for 2 h. The sections were then incubated in primary antibody overnight in a moist chamber at 4 °C. The grids were washed several times in PBS and incubated in gold tagged secondary antibody (anti-mouse IgG or anti-rabbit IgG conjugated with 20-nm colloidal gold particles; BB International/Ted Pella, Inc., Redding, CA) for 2 h at room temperature. Most sections were viewed without lead citrate-uranyl acetate counterstain, as this facilitated identification of gold particles. When stained, however, sections were incubated with aqueous uranyl acetate (10 min) and Reynold's lead citrate (4 min). Sections were examined using a Jeol 1200EX transmission electron microscope with an AMT Digital Imaging System.
For confocal microscopy, 5 consecutive fields (visualized with a 40× objective lens) were obtained beginning with the dorsal-most portion of Cg3 (3 fields) through the infralimbic area (2 fields). Sampling was performed separately within layers II and V. Each field was analyzed by first locating a GABA-positive neuron, then visualizing the fibers associated with that cell by examining consecutive optical planes through the respective images. Contacts between BLA fibers (red) and GABA-ir somata and dendrites (green) were marked separately for each optical plane. Data were expressed as the number of contacts per square micron of field (for each optical plane). The data were analyzed using bivariate regression analyses with the density of contacts as the dependent variable and age as the independent variable.
For the EM studies, cell bodies, and/or dendritic profiles with and without colloidal gold particles marking the presence of GABA were photographed. Axon terminations associated with fibers of the basolateral amygdala were identified by the presence of immunoperoxidase reaction product. All cell bodies and dendrites containing colloidal gold were photographed, regardless of whether there were BLA fibers in the field. The number of positive fibers forming synaptic (i.e., Gray Type I or Gray Type II) contacts, as well as other nonsynaptic appositions (i.e., without postsynaptic membrane specializations) with dendritic spines or shafts was determined for profiles with or without colloidal gold. An R × C contingency table analysis was used to assess whether the relative proportion of BLA fibers in contact with labeled or unlabeled spines or shafts was different.
As with previous studies in this (Cunningham et al. 2002) and other laboratories (Krettek and Price 1977; Kita and Kitai 1990; Bacon et al. 1996; Verwer et al. 1996; Cunningham et al. 2002), injections of anterograde tracer into the BLA resulted in a bilaminar distribution of amygdalofugal fibers forming dense bands in layers II and V of the mPFC. A double immunoreaction for light microscopy was utilized to gain an initial view of the interaction of BLA fibers with GABAergic interneurons. Figure 1 illustrates the elaborate BLA fiber network within layer V, as well as its distribution with respect to GABAergic neurons. Careful microscopic observation revealed that essentially every GABAergic neuron appeared to receive at least one contact from a BLA fiber. These interactions were characterized by the frequent presence of a swelling, or varicosity, at the point of contact. Both somatic and dendritic interactions were clearly observed.
Because of the limited resolution of brightfield microscopy, laser scanning confocal microscopy was used to confirm and quantify actual contacts of BLA fibers with GABA-ir elements in layers II and V of the CG3 region using a double fluorescent approach. The present studies, first of all, corroborated a previous report that the density of GABAergic neurons within the mPFC is maximal in the rat at approximately P5 (Vincent et al. 1995), but shows a negative relationship to the thickness of the cortical mantle. At approximately P20, the numerical density of these neurons shows a sharp decline as the relative amount of neuropil surrounding these cells expands, and as axonic and dendritic fibers from intrinsic and extrinsic sources increase (Vincent et al. 1995).
Figure 2 illustrates the decrease in numerical density of GABAergic neurons observed between P6 and P25, as well as the progressive increase in amygdalo-cortical fibers up to P56 (Fig. 2A–D). High magnification analysis of confocal optical sections in the z-axis demonstrated that BLA fibers are indeed in contact with GABAergic elements, as they are covisualized within the same 0.5 μm optical sections (Fig. 3B–E).
BLA appositions with cell bodies (“axosomatic”) and with dendrites (“axo-dendritic”) within the CG3 region were quantified for the preweanling (P6 and P16), early postweanling (P25 and P35), late postweanling/early adult (P45 and P56), and adult (P120) periods. As shown in Fig. 4, bivariate plots of the Bioquant data demonstrated an 8-fold curvilinear increase in axo-dendritic contacts in both layers II (r = 0.921) and V (r = 0.962) and a 3-fold increase in axosomatic contacts in both layers II (r = 0.712) and V (r = 0.771).
Although actual appositions were demonstrated with confocal microscopic analysis, ultrastructural characterization was needed to establish whether there might be interposed nonneuronal structures lying between the boutons and the cell surface and/or whether they might be synaptic in nature. Using a peroxidase reaction for BLA fibers and a colloidal gold immunoreaction against GABAergic elements, amygdalofugal axons were identified by the presence of a diffuse electron-dense reaction product, whereas GABAergic elements were identified by the presence of 20-nm colloidal gold particles (Figs 5 and 6).
Figure 5A depicts a GABAergic neuron labeled with numerous colloidal gold particles and juxtaposed with an incoming BLA fiber. Note the absence of colloidal gold in the surrounding parenchyma indicating exquisite specificity of the labeling method. At higher power, the peroxidase-labeled fiber is observed in close apposition with the neuron's dendrite, although no membrane specialization is appreciated in this region of contact. This fiber is, however, associated with a synaptic specialization at its apposition with another element. Figure 5B illustrates the frequently observed “axo-spinous” contact, with the postsynaptic spines typically being nonreactive for GABA (see Table 1). These data are consistent with our previous study that demonstrated that approximately 33% of BLA fibers form contacts with dendritic spines (Cunningham et al. 2002). It is noteworthy that postsynaptic membrane specializations were consistently observed on GABA-negative spinous processes in apposition with BLA fibers. Based on general observations of cortical neuropil, at least 90% of the latter structures are probably associated with pyramidal neurons, because most interneurons have aspiny dendrites. Figure 5C demonstrates an apposition of a BLA fiber with a dendritic shaft (axo-dendritic) containing colloidal gold immunoreactivity for GABA, and this was the most commonly observed interaction (36.5%) of amygdalofugal fibers with mPFC elements (Table 1).
|GABA+||19 (36.5%)||4 (7.7%)||3 (5.8%)|
|GABA−||9 (17.3%)||17 (30.8%)||1 (1.9%)|
|GABA+||19 (36.5%)||4 (7.7%)||3 (5.8%)|
|GABA−||9 (17.3%)||17 (30.8%)||1 (1.9%)|
Note: Total number of profiles from a set of electron photomicrographs showing appositions of anterogradely labeled BLA fibers with neuronal elements identified as immunoreactive for GABA (GABA+) or nonreactive (GABA−). The highest percentage of BLA interactions was seen with GABA+ dendritic shafts.
A semiquantitative analysis was performed to determine whether postsynaptic membrane specializations like those seen in Gray type I (excitatory) or Gray type II (inhibitory) synapses were present in GABA-positive versus GABA-negative elements (Gray 1959). Type I synapses are typically found on dendrites, particularly spines, and show an asymmetric membrane specialization that is almost exclusively found on dendritic spines. Type II synapses are primarily found on the cell bodies of neurons and/or proximal dendritic branches and have a symmetric membrane specialization that is found on both the pre- and postsynaptic elements. Figure 6 illustrates the ultrastructural detail of a BLA contacts with GABA-positive dendrites shown at high magnification. We typically observed a BLA fiber interdigitating with a dendritic shaft protuberance, but we could not identify membrane densities or specializations that were synaptic in nature. However, in many cases, the same BLA fibers forming nonsynaptic appositions with GABAergic elements were observed interacting synaptically with other, non-GABAergic elements.
The experiments reported here have demonstrated that GABAergic interneurons appear to be a primary target of sprouting amygdalar afferents projecting to rat mPFC where they form progressively greater numbers of contacts during the adolescent and early adulthood periods. This phase of rapid biopsychosocial growth is the time when many forms of psychopathology manifest, and suggests a general role for this connectivity in the maturation of limbic function. Ultrastructural observations have further revealed contacts with both axo-dendritic and axo-spinous contacts, the former occurring primarily with GABA-positive elements and the latter with GABA-negative elements. The majority of BLA interactions with spines showed classic Gray type I synapses having asymmetric membrane specializations, whereas contacts with GABA-positive dendritic shafts did not reveal discernable membrane specializations. This latter finding was unexpected, but could be explained in at least 3 different ways. First, membrane specializations at the point of contact between BLA fibers and GABA-positive dendrites may not exist and this interaction is nonsynaptic in nature. Secondly, membrane specializations at these sites might have been present prior to death; however, agonal changes and/or perfusion fixation could have promoted their degradation during processing of the tissue for the double EM immunolocalization. Third, because the length and diameter of dendritic branches are, generally speaking, much larger than those of spines, it is probably more difficult to detect synaptic contacts on shafts. For axo-spinous synapses, the dimensions of terminal bouton are similar to the spine heads and, in ultrathin sections, there is a much higher likelihood that membrane specializations will be observed. For BLA contacts on dendritic branches, only a serial EM analysis would be capable of providing a precise estimate of whether membrane specializations are present on the shafts of the dendritic tree that are in contact with BLA fibers (e.g. see Seguela et al. 1988).
BLA projection neurons are believed to be excitatory in nature (McDonald 1996, p. 48; Cunningham 2007, p. 181) and the inhibitory postsynaptic potentials that have been detected with in vivo recordings in rat mPFC imply that these fibers may induce a primary activation of GABAergic interneurons, followed by a secondary inhibition of principle cells (Perez-Jaranay and Vives 1991). Some investigators have suggested that fiber projections originating from the thalamus (Staiger et al. 1996), as well as the motor and sensory cortices (Ramanathan et al. 2002) form classical asymmetric synapses on inhibitory interneurons. In rat mPFC, single cell recordings from interneurons in layers II and V of rat mPFC would be needed to determine whether the response of cortical GABA cells to BLA activation is synaptic in nature.
GABAergic modulation of the corticolimbic system is pivotal to emotional, perceptual, and cognitive processing. The postnatal increase in the interaction of BLA fibers with GABAergic interneurons in the mPFC has implications for the development of the ability to process higher order information. Just as sensory experience in early life shapes the mammalian visual or primary sensory cortex (Feldman and Brecht 2005), emotional experience during development may shape the limbic-cortical relay system. The GABAergic neuron appears to be critical to the normal development and functioning of these networks. Mice deficient in the 65-kDa isoform of the GABA-synthesizing enzyme, glutamic acid decarboxylase (GAD65), do not undergo the synaptic modification in visual cortex that normally occurs with brief monocular deprivation. The loss of such activity-dependent plasticity can be reversed by restoring GABAergic balance with diazepam (Hensch et al. 1998). Presumably, a disturbance in the sprouting of BLA fibers in the mPFC during a critical stage of postnatal development would be associated with disruptions in the integration of emotional and perceptual information.
Schizophrenia is broadly believed to be a neurodevelopmental disorder in which GABA cell dysfunction in ACCx plays an important role (Benes 2000). The finding of abnormalities in the GABA system was foreshadowed by the observation that schizophrenics are unable to distinguish or “filter” out irrelevant stimuli from relevant information (McGhie and Chapman 1961). A plausible mechanism for GABA cell dysfunction in schizophrenia within amygdalar terminal fields is that abnormally increased amygdalar activity may produce an environment of increased glutamatergic transmission and possibly even oxidative stress (Coyle and Puttfarcken 1993). With the present data demonstrating increased targeting of GABAergic cells by BLA afferents primarily during the adolescent the early adulthood periods, it is possible that the normal ingrowth of projections from the amygdala may “trigger” the onset of schizophrenia and indeed other forms of psychopathology by influencing the expression of unique sets of susceptibility genes that define specific cellular endophenotypes within vulnerable subpopulations of GABA cells (Benes et al. 2007). It is also possible that the developmental plasticity of BLA fibers during the postnatal period could result in abnormal patterns of sprouting that result in “mis-wired” GABA cells within complex circuits. Based on the results reported here, such cellular and molecular mechanisms could be activated during childhood, adolescence, and early adulthood.
Overall, the results of this study extend those of an earlier report indicating that fibers from the BLA are actively sprouting within the mPFC during the postweanling period. The current findings now suggest that these amygdalar fibers are forming increased contacts with GABAergic interneurons as late as the early adult period. However, this connectivity requires further evaluation ultrastructurally, and the identity of the receptors at these appositions merits investigation. Further, a more detailed understanding of how these amygdalar fibers may influence specific subtypes of GABAergic interneurons, defined by their content of various calcium binding peptides, will be an important avenue to pursue. Previous work has suggested that there are differences in the response of parvalbumin-, calbindin-, calretinin-, and cholecystokinin-containing cells to BLA activation of the hippocampus (Berretta et al. 2004). It will be critical to identify whether such changes occur at the same or different stages of the postnatal period and how they influence the output of intrinsic circuits within the rat mPFC.
This work was funded by grants from the NIH (MH00423 and MH60450) and the Stanley Foundation.
Conflict of Interest: None declared.