To delineate the cellular mechanisms underlying the function of medial prefrontal cortex (mPFC) networks, it is critical to understand how synaptic inputs from various afferents are integrated and drive neuronal activity in this region. Using a newly developed slice preparation, we were able to identify a bundle of axons that contain extraneocortical fibers projecting to neurons in the prelimbic cortex. The anatomical origin and functional connectivity of the identified fiber bundle were probed by in vivo track tracing in combination with optic and whole-cell recordings of neurons in layers 2/3 and 5/6. We demonstrate that the identified bundle contains afferent fibers primarily from the ventral hippocampus but does not include contributions from the mediodorsal nucleus of the thalamus, amygdala, or lateral hypothalamus/medial forebrain bundle. Further, we provide evidence that activation of this fiber bundle results in patterned activity of neurons in the mPFC, which is distinct from that of laminar stimulation of either the deep layers 5/6 or the superficial layer 1. Evoked excitatory postsynaptic potentials are monosynaptic and glutamatergic and exhibit bidirectional changes in synaptic efficacy in response to physiologically relevant induction protocols. These data provide the necessary groundwork for the characterization of the hippocampal pathway projecting to the mPFC.
The medial prefrontal cortex (mPFC) has been viewed as a pivotal site for multiple cognitive functions. Lesion of the mPFC results in cognitive deficits that resemble symptoms of schizophrenia and Alzheimer's disease (Kolb 1984, 1990; Heckers et al. 1998). Emerging evidence suggests that in addition to dysfunction of a specific cortical area, failure of functional integration of multiple brain regions also contributes to the etiology of many disorders. Therefore, it is important to understand how synaptic inputs from various regions are integrated to drive neuronal activity of the mPFC.
Of particular importance to the proper functioning of mPFC during cognitive tasks is the afferents originating from the ventral hippocampus (vHipp) (Ferino et al. 1987). Previous work has shown that the CA1 and subicular subdivisions of vHipp send ipsilateral, unidirectional, and glutamatergic projections to mPFC (Jay et al. 1992, 1995). This hippocampal–mPFC pathway has been described as projecting anteriorly through the fornix. It ultimately ascends dorsally to terminate on neurons within the nucleus accumbens (NAcc) and a spatially restricted region of the mPFC that includes the infralimbic (IL) and prelimbic (PrL) cortices (Jay and Witter 1991; Jay et al. 1992). The functional integrity and flow of information between the hippocampus and mPFC are of critical importance to the proper functioning of the mPFC (O'Donnell and Grace 1995; Seamans et al. 1995; O'Donnell et al. 2002; Goto and O'Donnell 2003; Goto and Grace 2008). For example, simultaneous recordings in the hippocampus and mPFC of freely moving animals demonstrate that correlated activity in these 2 structures is phase locked at theta rhythm (Siapas et al. 2005). Interestingly, correlated firing activity between the hippocampus and mPFC is enhanced during specific behaviors that recruit spatial working memory (Jones and Wilson 2005a).
The rodent mPFC has been a good model for delineating cellular mechanisms underlying integration of polyassociational input for control of higher order cognitive functioning (Kolb 1990). Generally 2 approaches have been taken to obtain synaptic responses in cortical slices. First, 2 synaptically connected neurons are recorded simultaneously (Sjostrom et al. 2001, 2003, 2004; Gao and Goldman-Rakic 2003). Stimulating 1 neuron yields monosynaptic responses in the other. However, such paired recordings are limited to studying local circuits because the contribution of synaptic input from outside the mPFC is completely excluded. The other frequently used method is to place stimulation electrodes in different layers of the mPFC because pyramidal neurons have elaborate dendrites that extend into deep and superficial layers of the cortex (Matsuda et al. 2006). However, like other cortical regions, the mPFC contains local circuits and receives innervations of afferents that are not spatially segregated. Consequently, synaptic responses resulted from general laminar stimulation arise from a mixed population of axonal fibers and may include complex polysynaptic responses. The inability to activate specific inputs in a slice preparation represents a technological limitation and has prevented investigations into the unique nature of information processing by different synapses in the mPFC.
In this study, we developed a slice preparation with the goal of preserving the hippocampal afferent fibers projecting to the mPFC, especially to the PrL region. Using a combination of anatomical and electrophysiological approaches, we demonstrate that hippocampal afferent fibers can be identified outside of the mPFC. Stimulation of these fibers results in monosynaptic responses that undergo bidirectional plasticity in response to patterned activities. Thus, this new preparation holds great potential for studying detailed cellular properties and mechanisms that direct information flow from the hippocampus to the mPFC.
Materials and Methods
The use of animals for the studies described below was approved by the University of Minnesota Institutional Animal Care and Use Committee.
Preparation of Mouse Prefrontal Cortex Slices
Coronal slices containing mPFC were prepared from 8- to 15-week-old mice. Animals were anesthetized by lethal dose of a mix of ketamine and xylazine and perfused through the heart with ice-cold cutting solution containing the following (in millimolar): 240 sucrose, 2.5 KCl, 1.25 NaH2PO4, 25 NaHCO3, 0.5 CaCl2, and 7 MgCl2, saturated with 95% O2/5% CO2 before freezing. Both hemispheres were then quickly removed, and modified coronal slices containing PrL and IL regions were cut at 300-μm thickness using a Vibratome. After incubation in a holding chamber containing normal artificial cerebrospinal fluid (ACSF) for at least 30 min at room temperature (RT), slices were transferred into the recording chamber. A Zeiss Axioskop 2 FS, fitted with 40× water-immersion objective and differential interference contrast (DIC), was used to view slices. Light in the near infrared range (740 nm), in conjunction with a contrast-enhancing camera, was used to visualize individual dendrites. For all recordings, the bath solution (ACSF) contained (in millimolar) 125 NaCl, 2.5 KCl, 1.25 NaH2PO4, 25 NaHCO3, 2.0 CaCl2, 1.0 MgCl2, and 10 dextrose unless otherwise stated.
Electrophysiological Recordings and Focal Stimulations
Dagan 700A and Axopatch 200B amplifiers were used for current and voltage clamp recordings, respectively. Whole-cell recording pipettes (5–8 MΩ) contain (millimolar) 120 Kgluconate, 20 KCl, 10 4-(2-hydroxyethyl)-1-piperazineethanesulfonic acid, 0.2 ethyleneglycol-bis(2-aminoethylether)-N,N,N′,N′-tetraacetic acid, 2 Mg2Cl, 4 Na2ATP, 0.3 Tris–GTP, and 14 phosphocreatine (pH 7.25 with KOH). Biocytin (0.2–0.4%) was occasionally included for morphological confirmation. Pulse generation and data acquisition were controlled with custom software written in the Igor Pro environment. The hippocampal axonal bundle was stimulated electrically with a microelectrode placed within the bundle and controlled by a fine micromanipulator at the resolution of 1 μm. Detailed information on topographical arrangement of stimulating and recording electrodes is provided in Experimental Results. To induce long-term potentiation (LTP), we used a theta burst pairing protocol as described before (Watanabe et al. 2002). Briefly, 5 subthreshold excitatory postsynaptic potentials (EPSPs) at 100 Hz were paired with action potentials (APs) elicited with somatic depolarizing current injection into the mPFC pyramidal neurons. The pairing was repeated 80 times at 5 Hz. Long-term depression (LTD) was induced by EPSP and AP pairing repeated 900 times at 3 Hz (Christie et al. 1996). Test stimuli are delivered every 30 s; a hyperpolarizing current pulse was injected into the cell after the test stimulus to monitor input resistance and series resistance throughout recording. Slope measurements of EPSP were made from a line fitted to the rising phase of the EPSP. The magnitude of LTP or LTD was quantified as percentage of changes of the EPSP slope 25 min after induction.
In Vivo Injection of Anterograde Tracers
We employed an approach similar to that described previously for in vivo injections of various brain regions of mice (Wong et al. 1999). Adult male mice (8–10 weeks old) were anesthetized intraperitoneally with ketamine (90 mg/kg)/xylazine (10 mg/kg) and placed in a stereotaxic frame (Stoelting Instrument, Stoelting Co., Wood Dale, IL). Additional doses of anesthetic were administered periodically. Coordinates for targeting the brain regions were obtained from a mouse brain atlas and are summarized here (Paxinos and Franklin 2001). For vHipp CA1 region: bregma −3.88 mm, lateral: 3 mm, depth: 2.5 mm from the cortical surface; amygdala (bregma [B]: −1.6, lateral [L]: 3.2, depth [D]: −4.2); mediodorsal nucleus (MD) of thalamus (B: −1.2, L: 0.5, D: 2.9–3); and lateral hypothalamus (LH) (B: +0.6, L: −0.9, D: 5.6). The dura mater was carefully removed to expose the brain surface. Quartz micropipettes with a filament (Sutter Instruments, Invitrogen, Carlsbad, CA) were beveled to a tip size of 5–10 μm. Microliter injections of Texas Red– or fluorescein-conjugated dextran anterograde tracer (Invitrogen) were made using either pressure injection via a picospritzer or current pulse injection through the micropipette. Mice were allowed to recover for 4–15 days before euthanasia. The brains of injected mice were cut into slices and fixed (4% paraformaldehyde) prior to being mounted onto glass slides.
Calcium Imaging and Analysis
Three- to five-week-old mice were used for Ca2+ imaging experiments. To load the calcium-sensitive AM-ester dyes Calcium Green Fura-2 or Oregon Green, 7 μL of 1 mM dye and 1.4% Pluronic F-127 in dimethyl sulfoxide were applied directly to the slice bathed in 2.5 mL ACSF aerated with 95%/5% O2/CO2 at RT (MacLean and Yuste 2005). After 20–60 min staining, slices were placed directly into the recording chamber and continuously bathed in 34 °C circulating ACSF bubbled with 95%/5% O2/CO2. Calcium imaging was carried out with a Red Shirt imaging/Neuroplex SM256 system on an Olympus BX51WI microscope. Slices were imaged using a low-magnification, high-aperture objective (20×, numerical aperture [NA] = 0.95) with individual frames of 256 × 256 pixels. Imaging frames were acquired for 100 ms prior to stimulation and for up to 4 s poststimulation at 100 Hz. Stimulation was carried out using a Grass stimulator providing a stimulation pulse of 100–150 μA for 0.1 ms.
Imaging data were analyzed using custom code written in MATLAB. Change in fluorescence was calculated as ΔF/F by subtracting each poststimulation imaging frame from a prestimulation baseline reference frame divided by the baseline reference frame. (ΔF/F = [poststimulation frame − baseline reference frame]/baseline reference frame.) When imaging adjacent spatial domains on a single slice to reconstruct the neuronal activity along the entire mPFC slice surface, each constituent frame's baseline level of fluorescence was set to the prestimulation baseline level. The quantification of fluorescence changes across the mPFC spatial domain was calculated by a linear sum of all pixel values across a single dimension of the image (horizontal axis or vertical axis). Comparisons of fluorescence change were made within a single-image frame or within combined frames of multiple mPFC spatial domains.
Development of a mPFC Slice Preparation where Extracortical Fiber Bundles Can Be Identified
Tracing studies describe the hippocampus CA1/subiculum (Sub) pathway to the mPFC as coursing through the fimbria and fornix to reach the medial portion of the lateral septum, NAcc, and the PrL and IL cortices (Jay and Witter 1991; Thierry et al. 2000; Hoover and Vertes 2007). On the basis of these observations, we made efforts to develop a modified coronal slice preparation of mouse forebrain that would capture the transition point where the ventrally projecting hippocampal fibers leaving the fornix turn dorsally to innervate the NAcc and mPFC (Fig. 1A). The ideal slice with the highest probability of retaining functional connectivity was found rostral to the crossing of the anterior commissure and caudal enough within the prefrontal cortex (PFC) to contain the most rostral portion of the cingulum and the lateral ventricle (Fig. 1B). The coronal slices used throughout experiments were cut at an angle of ∼10 to 12° caudal from a tangential axis through the dorsal plane of the brain. These slices contained the most posterior PrL and IL cortices as well as dorsal tenia tecta (DTT) according to the mouse brain atlas of Paxinos and Franklin (2001).
When looking along the ventromedial surface of the PFC slice under bright-field microscopy, various boundaries can be identified from the midsagittal fissure to the lateral ventricle/NAcc area (Fig. 1B, right panel). Most evident, at a level ventral to cingulum were well-identified layer 1 and layer 2/3 (Fig. 1B, position 1 and 2, respectively), a medial (3) and more lateral (4) portion of a large fiber bundle system, an unidentified region of white matter between the fiber bundle system and the lateral ventricle (Fig. 1B, position 5), and finally the lateral ventricle (Fig. 1B, position 6) and NAcc. Although axons in the medial aspect of the fiber bundle system were cut, fibers that lay in the more lateral portion of this area appeared to be intact (position 5). These intact fibers paralleled the medial fiber system and followed the cingulum along the dorsoventral extent of this slice preparation. Under the DIC microscope, it was possible to identify what looked like intact axonal fibers traveling directly between DTT and the lateral ventricle to more dorsal areas of the frontal pole. We hypothesized that this unidentified white matter region contained fibers that originated from the hippocampus and projected to the mPFC.
Characterization of Excitatory Responses in the mPFC by Extracortical Stimulation
Using monopolar electrodes positioned along the presumable hippocampal afferents (Fig. 2A, open circles), we were able to deliver focal stimulations and obtain EPSPs in neurons located in the PrL region of mPFC (Fig. 2B). The distance between the stimulation and recording sites was usually 550–700 μm. EPSPs recorded in layer 5 pyramidal neurons exhibited a short delay between stimulation and the onset of the EPSP (3.98 ± 0.09 ms, n = 101). This delay was short enough to exclude polysynaptic responses. A binned histogram of delay times recorded from all neurons was best fit with a Guassian curve, which exhibited a normal but not bimodal distribution (Fig. 2D). EPSPs recorded from layer 2/3 pyramidal neurons showed similar synaptic delay (4.18 ± 0.06 ms, n = 208). Synaptic delays for a given cell remained consistent at around 4–5 ms when we manipulated the stimulation intensity while maintaining a constant distance between the recorded neuron and the stimulation site. Calculation of the conduction velocity of APs was based on placing stimulation at 2 different positions along the hippocampal bundles while recording EPSPs in the same postsynaptic neuron. The conduction velocity measured in this in vitro preparation as the distance between stimulation electrodes divided by the difference of synaptic delay (0.18 ± 0.05 m/s, n = 3) is in the same order of velocity measured in vivo (0.6 m/s) (Ferino et al. 1987; Jay et al. 1992).
To rule out the possibility that the fiber bundles we stimulated were projecting axons leaving the mPFC, we investigated the probability of obtaining antidromic APs under different conditions. In several hundred cases, we never observed a single antidromically induced AP when the stimulus electrode was positioned at the region where the presumed hippocampal afferents are located, >550 μm from the recorded neuron (Fig. 2A, open circles). As a control, we intentionally targeted the exiting axons located between PrL layer 6 and the cingulum as well as regions in the cingulum (Fig. 2A, filled circles), according to the previous descriptions of the mPFC efferent fiber distribution (Sesack et al. 1989). In the presence of glutamate receptor blockers 6-Cyano-7-nitroquinoxaline-2,3-dione (CNQX) (20 μM) and 2-amino-phosphonovaleric acid (APV) (50 μM), we recorded antidromic APs with high probability (7/7 cells, Fig. 2C). The sharp contrast in encountering antidromic APs between these 2 conditions strongly suggests that the identified fiber bundle does not contain mPFC efferents.
Anterograde Tracing of Identified Fiber Bundles In Vivo
With the above physiological recordings indicating monosynaptic connections, we set out to test our hypothesis that the longitudinal fibers projecting to the PrL cortex in the slice preparation originated from the vHipp and Sub (vHipp/Sub). In addition to vHipp/Sub, mPFC also receives inputs from cortical and other subcortical regions. A series of retrograde and anterograde track-tracing studies suggest that major noncortical regions projecting to the mPFC include the vHipp, the MD of the thalamus, the amygdala, and the LH/medial forebrain bundle (Saper 1985; Sesack et al. 1989; Hoover and Vertes 2007). However, information on the spatial relationship between these projections as they course toward the mPFC is not explicitly described; yet, these details are necessary for designing slice preparations that preserve specific afferent fibers.
Therefore, in this study, we have carried out direct iontophoretic injection of dextran anterograde tracers into the vHipp (n = 8), amygdala (n = 3), MD (n = 2), and LH (n = 2) to determine the spatial trajectories of major afferent fibers into the mPFC.
Fluorescently labeled projection fibers exited the vHipp via the alveus proximal to the site of tracer injection (Fig. 3A). These fibers coursed around the hippocampus to join the fimbria. Labeled fibers were found in the dorsal aspect of the fimbria and proceeded rostrally through the brain via the fornix system. Upon reaching the columns of the fornix, fibers coursed rostroventrally toward the anterior commissure. Rather than continuing toward the mammillary bodies in a posteroventral direction similar to the classical route of fibers within the columns of the fornix, hippocampal efferent fibers destined to innervate the mPFC proceeded anteroventrally toward the medial portion of the ventral striatum (Fig. 3B). These data confirmed previous observations by others (Jay and Witter 1991) and provided basic anatomic knowledge for the proper targeting of the tracer into the vHipp/Sub. Furthermore, labeled fiber bundles turned dorsally and ran medial to the striatum yet lateral to the adjacent DTT. At this rostral position within the mouse brain, hippocampal efferent fibers are longitudinally parallel to, or slightly deviated from, the coronal plane. In the slice preparation designed to preserve these hippocampal axons, labeled fibers were indeed present directly within the region of stimulation (Fig. 3C, arrow), specifically coursing dorsally between the deepest layers of the mPFC and the cingulum prior to innervating their target mPFC laminations.
MD of the Thalamus
Following injection of the anterograde tracer into the MD, labeled efferent fibers en route to mPFC traversed the thalamus via a series of anteroventral radiations that extended laterally and entered the internal capsule (Fig. 4A). Labeled fibers from the MD retained a fasciculated appearance as they traveled through the dorsal striatum and finally crossed through the corpus callosum to innervate the overlying cortex. Specifically, we observed strong, positive signals in layer 2/3 of both PrL cortex and the orbital frontal cortex (OFC) (Fig. 4G, arrow). In addition, cell bodies of layer 5/6 neurons in the OFC were also labeled (Fig. 4G, arrowhead). Because one of the original defining characteristics of the PFC is its reciprocal innervation with the MD (Krettek and Price 1977; Gabbott et al. 2005), projecting neurons in the OFC may be labeled retrogradely by dye injected into the MD. Although slices containing the OFC were more anterior than our slice preparation, it was clear that fibers from the MD project to the frontal cortex but do not pass the region of stimulus electrode placement in our slice preparation (Fig. 4B).
The basolateral and lateral nuclei of the amygdaloid complex have been identified as major areas that project to both PrL and IL cortices (Kita and Kitai 1990; Bacon et al. 1996; Verwer et al. 1996). We have consistently targeted the fluorescent tracer into these regions (Fig. 4C, arrow). Following the major fibers exiting the amygdala and projecting anteriorly, we found labeled fiber pathways that innervated the ventral striatum initially coursed dorsally through the brain, medial to the dorsal striatum (caudate /putamen), and entered into the stria terminals (Fig. 4H). At this caudal level, labeled fibers in the stria terminalis were found to embrace the medial surface of the dorsal striatum as the fibers move rostrally toward more anterior structures in the stereotypical manner described in previous reports (Kita and Kitai 1990). However, we found no signal present in the stimulation region of the slice preparation (Fig. 4D). It seems that none of the amygdala pathways pass through the region of our interest.
Although it has been shown that the hypothalamus only provides a limited afferent input to the mPFC (Hoover and Vertes 2007), the LH is characteristic of bed nuclei in the brain. The LH contributes to the medial forebrain bundle which itself courses throughout the extent of LH (Nieuwenhuys et al. 1982). Furthermore, the medial forebrain bundle contains a variety of fibers including monoaminergic and cholinergic fibers that originate from various nuclei in the midbrain and brain stem, such as the ventral tegmental area, the substantia nigra, the dorsal and medial raphe, the locus coeruleus, as well as efferent LH fibers (Nieuwenhuys et al. 1982). Thus, it is very possible that tracer injected into the LH labeled nonglutamatergic fibers passing through this region.
We injected tracer into the LH and found positive but weak signals in the region of interest (Fig. 4E,F). However, unlike the hippocampal efferents that form a bundle in this region, fluorescently labeled axons from the LH were very sparse (Fig. 4I, arrow). Furthermore, these sparely labeled fibers were spatially segregated from those arising from the vHipp (Fig. 4F,I, asterisk). Specifically, they were located slightly medial to the hippocampal axon bundle.
In summary, our track-tracing results, in combination with previous anatomic descriptions, support the notion that an extracortically located fiber bundle contains axons from the vHipp but not the MD, amygdala, LH, or fibers passing through LH via the medial forebrain bundle.
Activation of Hippocampal Afferents Resulted in Patterned Activity in the mPFC
Recording synaptic responses from individual neurons in the mPFC demonstrates the connectivity of these neurons with the identified hippocampal afferent fibers. However, these data, obtained from recording a single neuron a time, lack information on the spatial distribution of mPFC network activity that results from hippocampal afferent stimulation. In order to specific spatial information, we needed to record multiple neurons simultaneously. To this end, we chose to monitor Ca2+ responses as an index of the synaptic/neuronal activity in a population of neurons (Fig. 5). This approach has been used for a similar purpose in the visual cortex (MacLean et al. 2006). The stimulus electrode was positioned at 3 locations, representing different fibers projecting to the mPFC. First, we stimulated the identified hippocampal afferents at >550 μm ventral to the dorsal aspect of the cingulum (Fig. 5A), as described earlier. Neurons surrounding the stimulus electrode became activated due to activation of local network. Imaging within the PrL and dorsal IL cortices demonstrated that neurons located distal from the point of stimulation exhibited a laminated pattern of activation that was restricted to layer 5/6 without extending laterally (n = 7). The punctate pattern of activated neurons may represent single or clustered groups of neurons that are activated by stimulating hippocampal axons. Within the superficial layers of the PrL cortex (medial to the deep layers), there was a minimal level of activation with the exception of few neurons located in layer 2/3.
Next, we conducted similar experiments by placing the stimulus electrode in layer 5/6 or layer 1 of the mPFC. Stimulation of layer 5/6 only activated neurons surrounding the stimulation electrode indiscriminately without the obvious punctate patterns (Fig. 5D). Change in fluorescence declined in all directions to ∼20% of the peak signal at 120 μm from the stimulation site (n = 2). Alternatively, stimulation of layer 1 yielded robust neural activation that extended across all layers down to the most lateral aspect of layer 5/6 (Fig. 5E, n = 2). We observed a mixture of diffuse and punctate fluorescent signal patterns that were not restricted to a specific layer. Both layer 5/6 and layer 1 stimulation exhibited characteristics that indicate the gross involvement of local networks.
Glutamatergic Afferents Targeted Multiple Cell Types at Various Layers of the mPFC
As a first step in physiologically characterizing the hippocampal connections in vitro, we aimed at identifying targets of the hippocampal–mPFC projection. We made whole-cell recordings on various types of neurons based on the shape of the soma, the location, and the firing properties (Fig. 6A2,B2). A specific cell type was then confirmed by biocytin staining and subsequent morphological identification (Fig. 6A1,B1). In response to electrical stimulation of the visually identified axonal bundles, both pyramidal neurons and interneurons showed evoked EPSPs. The short synaptic delay excluded polysynaptic connections (Fig. 6A3,B3). We conclude that this potential hippocampal–mPFC afferent makes monosynaptic connections with both pyramidal neurons and interneurons in multiple regions of the mPFC (L2/3 vs. L5). Furthermore, coapplication of CNQX (20 μM) and APV (50 μM) completely abolished the evoked responses (Fig. 6B3), suggesting that this pathway is glutamatergic.
We also measured paired pulse facilitation (PPF) ratios (second Excitatory postsynaptic current [EPSC]/first EPSC) at 20-, 50-, 100-, 150-, and 200-ms intervals (Fig. 6C). Across all intervals, the PPF ratios of the hippocampal–mPFC pathway were significantly larger than 1, indicating a greater degree of facilitated presynaptic release. We then conducted the same experiments using layer 1 stimulation, where only fibers from other cortical regions are present. The PPF ratio profile seemed to be quite different, with the ratio being around 1 at most of the intervals except for 20 and 40 ms. It appears that PPF ratios of the hippocampal–mPFC pathway are significantly higher than that of the cortical–mPFC pathways.
Stability and Plasticity of Hippocampal Afferent Fibers In Vitro
In the slice preparation, we monitored EPSPs in pyramidal neurons (layer 2/3 and layer 5) in response to stimulation of hippocampal afferents for up to 100 min (n = 10). Both the slope and amplitude of EPSPs remained stable during this period of time (Fig. 6D), validating the use of the slice preparation and extracortical stimulation for future in vitro studies.
We induced LTP in the hippocampal–mPFC using a pairing protocol at 5 Hz (theta), a firing frequency detected in both hippocampus and mPFC in vivo when these 2 structures are coactivated (Jones and Wilson 2005b; Siapas et al. 2005). A theta burst pairing protocol involves correlated pre- and postsynaptic activities. Specifically, 5 subthreshold EPSPs at 100 Hz were paired with APs elicited with somatic depolarizing current injection into the mPFC pyramidal neurons. The pairing was repeated 80 times at 5 Hz (Fig. 6A,B). When EPSPs and APs were paired at a time window of 20–35 ms, synaptic efficacy of the hippocampal–mPFC pathway, measured as EPSP slope, was increased 30 min postinduction for both layer 2/3 (148 ± 13.4%, Fig. 6C, n = 7) and layer 5 neurons (156 ± 30%, n = 4). Perfusion of N-methyl D-aspartate receptor blocker APV (50 μM) during the induction period was able to prevent LTP (n = 3). These results are particularly important in that induction of LTP in mPFC pyramidal neurons has been established mostly by including γ-aminobutyric acidA receptor blockers in previous in vitro studies (Jay et al. 1996; Matsuda et al. 2006).
In addition to potentiation, this pathway underwent depression in response to a protocol consisting of 900 pairings of a single AP and EPSP at 3 Hz (Fig. 7D). Specifically, when we recorded pyramidal neurons in layer 2/3 and stimulated the hippocampal afferents, the synaptic efficacy was reduced to 58.8 ± 8% (Fig. 7E, n = 8). Collectively, these data indicate that the hippocampal–mPFC pathway exhibits bidirectional, long-term changes in response to physiologically relevant stimulations.
We described a modified coronal slice of mPFC on which an extracortically located fiber bundle can be visually identified. Stimulation of these fibers at a distance from the PrL region of the mPFC to exclude contamination from the local circuitry results in monosynaptic responses. Further track-tracing results support that this fiber bundle contains axons primarily from the vHipp but not other major mPFC-projecting regions including the MD, amygdala, and LH. Although most of what we know about the hippocampal pathway projecting to the PFC comes from in vivo experiments, this carefully designed in vitro preparation holds promise for studying in great detail the synaptic properties and plasticity mechanisms pertaining to this important pathway.
Validation of the Tracing Approach Employed
To probe the origin of the fiber bundle identified in the slice preparation, we injected fluorescently conjugated dextran into multiple brain regions that are believed to project to the mPFC. These regions include the vHipp, LH/medial forebrain bundle, amygdala, and the MD. In previous track-tracing studies, biotinylated dextrans or Phaseolus vulgaris–leucoagglutinin were used (Saper 1985; Jay and Witter 1991; Gabbott et al. 2005; Hoover and Vertes 2007). These tracers require post hoc processing to amplify signals and visualize the terminal distributions of afferent inputs. In this study, we were primarily interested in demarking the relative spatial profiles of afferent bundles as they project through the brain from their origins to the final destination in the mPFC. Moreover, the advantage of avoiding tissue postprocessing makes it possible to combine in vivo tracing and in vitro recording. Due to the primary focus of previous tracing studies on the synaptic terminal distribution of mPFC afferents, gaps exist in the fine details on how fibers move out of their major fiber bundles (fornix, stria terminals, internal capsule, and medial forebrain bundle) and into their regions of innervations. Our data provide detailed information on how mPFC afferent fibers project through the brain in a characteristic manner, leading to the successful design of a slice preparation that allows for identification and activation of the hippocampal projection.
Hippocampal Specificity of the Identified Fiber Bundle
In addition to other neocortical regions, it is generally agreed that the mPFC-projecting regions include hippocampus, amygdala, the midline thalamus, the medial basal forebrain, and monoaminergic nuclei of the brain stem (Saper 1985; Conde et al. 1995; Hoover and Vertes 2007). We injected tracers into 4 of these regions and obtained a positive signal in the region of stimulation only following injections of anterograde tracer into hippocampus. How homogeneous in composition is the fiber bundle that we stimulated in vitro? We cannot rule out that brain regions not picked up by previous retrograde tracing studies but possibly projecting to the mPFC may contribute to the fiber bundle we identified. Furthermore, in addition to the areas considered as major contributors projecting to mPFC, many small nuclei scattered throughout the brain may send a relatively small number of projection fibers to the mPFC that converge with the hippocampal input. However, our hypothesis of a primary hippocampal origin and the lack of a second large contributor to the fiber bundle is a reasonable interpretation of the data from this study.
Given the relative large size of the fiber bundle identified under DIC imaging and the spatial isolation of the stimulation region that elicits postsynaptic responses, it is possible that fibers from other brain regions and those projecting to neocortex are present along with this fiber pathway. However, this is irrelevant to recordings that require functional connections between presynaptic fibers and postsynaptic neurons. Fibers from other regions may become activated by stimulations, but they do not project to or are relatively sparse innervators of the mPFC. Thus, it is likely that they do not contribute, or contribute a relatively negligible amount, to the cellular responses of neurons located in the mPFC.
Another valid concern is whether the fiber bundles we stimulated are the axons of mPFC projection neurons rather than the axons from hippocampus. We believe that this is unlikely on the basis of probability of obtaining antidromic APs. Antidromically induced APs were obtained with high probability in expected regions of mPFC efferents (Sesack et al. 1989) but entirely absent in the region of hippocampal afferents among several hundred cases so far. This observation indicates that axons projecting from the mPFC either do not pass through the region where we identify the hippocampal afferents or are severed before reaching this region.
Potential Use of This Preparation
Two established in vitro slice preparations containing the thalamocortical projection have been valuable for physiological studies of this pathway (Agmon and Connors 1991; Cruikshank et al. 2002). Recent effort has aimed at developing preparations for studying other inputs into cortex, such as amygdala to mPFC (Orozco-Cabal et al. 2006) and lateral geniculate nucleus to the visual cortex (MacLean et al. 2006). Our success in preserving the hippocampal–mPFC connection in vitro provides opportunities for extensive studies of these synapses.
Cellular work on the hippocampal–mPFC pathway began with in vivo extracellular recordings of mPFC neurons following stimulation of the vHipp. These studies demonstrate that hippocampal synapses are glutamatergic (Thierry et al. 2000) and have the capacity to undergo potentiation (Laroche et al. 1990; Jay et al. 1995), depression (Takita et al. 1999), and depotentiation of previously potentiated synapses (Burette et al. 1997). However, in vivo experiments are somewhat restrictive and do not allow the full characterization of cellular mechanisms. An in vitro preparation allows access to regions and cellular manipulations that are not possible with in vivo work. This increased accessibility has provided for some unique observations in this study. We found that hippocampal–mPFC synapses exhibit higher PPF ratios than cortical inputs. Furthermore, synaptic efficacy undergoes bidirectional plasticity in response to physiologically relevant stimulation patterns, which is particularly relevant to specific behavioral phenomena. For example, correlated activity at theta frequency (5 Hz) between the hippocampus and the mPFC is enhanced during specific behaviors that recruit spatial working memory (Jones and Wilson 2005a, 2005b) but is decreased or abolished during error performance as reported by Hyman et al., Society for Neuroscience (2007). It would be of great interest to test whether certain neurotransmitters or modulators exert input-specific effects on the basal transmission and/or plasticity of this pathway relative to local networks. These differences may exist in presynaptic terminals of hippocampal afferents.
Finally, we have observed that in addition to those fibers projecting to the mPFC, a large portion of labeled hippocampal fibers turn ventrally and enter directly into NAcc in this slice preparation. Given the involvement of all 3 regions, the mPFC, NAcc, and hippocampus, in reward and addiction processes, further characterizations of these pathways would make this in vitro preparation an appealing model for studying multipathway integration and interactions.
National Institutes of Health (NS049129).
We thank Drs Walter Low and Terrence Burns for help with in vivo tracer injections, Drs Glenn Giesler and Esam El-Fakahany for comments on the manuscript, and Amber Lockridge and Brett Bennett for biocytin staining. Conflict of Interest: None declared.