We examined the extent of the ferret prefrontal cortex (PFC) and its reciprocal connections with the mediodorsal nucleus of the thalamus (MD) by anterograde and retrograde labeling in 6- to 14-week-old male ferrets. Our results indicate that in the ferret, as in other species, MD projects heavily to the PFC although it also projects to other cortical and subcortical structures. The MD projection to PFC terminates largely in layer IV with lighter innervation of layers II, III, V, and VI. The cells projecting back to MD are mostly in layer VI. The parvocellular component of MD projects to and receives projections from the more caudal and dorsomedial component of the PFC, whereas the magnocellular portion of MD projects to and receives projections from the more rostral and lateral component of the PFC. With these results we have localized the ferret PFC, defined as a frontal cortical region with heavy reciprocal connections with the MD.
The prefrontal cortex (PFC) is critically involved in the mediation of higher cognitive functions (Jacobsen 1936; Milner 1963; Goldman-Rakic 1987). Its integrity is compromised in disease, most notably in the neuropathology of schizophrenia whose characteristics include well-documented difficulties in performing tasks that entail skills for future planning, problem solving, working memory, and abstract reasoning (Franzen and Ingvar 1975; Weinberger et al. 1986; Goldman-Rakic 1994; Goldman-Rakic and Selemon 1997; Bunney and Bunney 2000).
For several decades, in vivo experiments carried out primarily in monkey PFC involved the correlation of single and/or multiple unit physiology with behavior in oculomotor delayed-response tasks designed to measure memory functions (Fuster and Alexander 1971; Fuster 1973; Funahashi et al. 1989; Compte et al. 2000; Constantinidis et al. 2001). Now, many laboratories, including ours, are increasingly using ferrets for in vivo and in vitro PFC functional microcircuitry studies (Gao et al. 2001; Krimer and Goldman-Rakic 2001; McCormick et al. 2003; Fritz et al. 2004, 2007; Haider et al. 2006; Shu et al. 2006, 2007) that address questions regarding neuronal interactions that cannot be approached reasonably in nonhuman primates. However, to our knowledge there are no accounts in the literature examining the extent, characteristics, and limits of the ferret PFC. Because those aspects of anatomy cannot be deciphered by in vitro studies, we undertook the task of investigating the extent of the ferret PFC using a hodological approach. The localization of the PFC of the ferret and the characterization of its extent are of fundamental importance for subsequent in vivo investigations in this species. Because of the limitations of in vitro and/or in vivo studies in large animals, it is advantageous to extend the investigation of PFC to other in vivo models. In addition to the ferret's ease of maintenance and mild, inquisitive, and altricial nature, significant advances have been made in the study of the development, organization, and physiology of the ferret brain (King et al. 1988, 1996, 1998; Moore et al. 1989, 1999; Voigt 1989; Voigt and de Lima 1991a, 1991b; Chapman and Stryker 1993; Voigt et al. 1993; Chapman et al. 1996; Durack and Katz 1996; Juliano et al. 1996; Noctor et al. 1997, 1999; Gao et al. 1999; Anderson et al. 2002; Innocenti et al. 2002; Manger, Masiello, et al. 2002; Manger, Kiper, et al. 2002; Restrepo et al. 2002; Bizley et al. 2007).
The main purpose of the present study was to characterize the ferret PFC by examining its heavy reciprocal connections with the mediodorsal nucleus of the thalamus (MD) using retrograde and anterograde tracers, as has been previously done in other species (Rose and Woolsey 1948; Tobias 1975; Markowitsch et al. 1978, 1980; Markowitsch and Pritzel 1979, 1981). It is hypothesized that projections from the parvocellular (lateral) and the magnocellular (medial) portions of MD target different neuronal populations in the PFC and that those neurons in turn project back selectively to their MD portions of origin. This study is also aimed at establishing the ferret as an appropriate model for in vivo studies of PFC that may be too costly and cumbersome to be performed in primates or other large animals. Preliminary results have been published in abstract form (Duque and Goldman-Rakic 2003).
Materials and Methods
Animals and Injections
Male ferrets (Mustela putoris furo) 6–14 weeks of age were used for this study. This age and gender are commonly used in experimental studies, including those in our own laboratory. To identify the zone of frontal cortex that receives a direct projection from the MD, biotin dextran amine (BDA, Molecular Probes, Eugene, OR) was injected into the MD (17 injections in 13 ferrets; 4 bilateral and 9 unilateral). As a control, BDA (4 injections in 3 animals) and wheat germ agglutinin conjugated horseradish peroxidase (WGA–HRP, Sigma, St Louis, MO) (2 injections in 2 animals) were injected into areas close to the MD. As an additional control, in one animal fluorogold (FG, Fluorochrome Inc., Denver, CO) and fast blue (FB) were injected ipsilaterally so that one tracer was in the MD while the other was not.
In order to perform retrograde tracing of cells that project to the MD-recipient zone of the PFC, BDA was injected into this cortical region (22 injections in 15 ferrets; 6 bilateral, 8 unilateral, 1 double ipsilateral). As a control, BDA (2 injections), WGA–HRP (2 injections), and FG (1 injection) and FB (1 injection) were injected into cortical areas adjacent to, but outside, the PFC (see Supplementary Table 1).
All procedures conformed to guidelines recommended in “Preparation and Maintenance of Higher Mammals During Neuroscience Experiments” (NIH publication No. 91-3207). All protocols were reviewed and approved by the Yale University Institutional Animal Care and Use Committee. Ferrets were purchased from Marshall Farms (North Rose, NY).
Surgeries were performed under aseptic conditions. Animals were anesthetized with a mixture of ketamine (20–30 mg/kg, im) and xylazine (1 mg/kg, im supplemented as necessary) and positioned in a small rodent stereotaxic apparatus fitted with a ferret adaptor (Kopf Instruments, Tujunga, CA). Ferrets were injected with atropine (0.04 mg/kg, im) and their eyes protected with ophthalmic ointment. Ear bars and the incisive foramen bar were infiltrated with xylocaine ointment (5%). Temperature was monitored and maintained between 38 and 39 °C with a water circulating heating pad.
Small burr holes (approximately 1 mm in diameter) were drilled above MD or the PFC. Dura matter was punctured and a 0.5- to 1-μL Hamilton syringe was used to deliver the tracer. The coordinates for MD and PFC were previously measured taking Bregma as a reference point in a custom made atlas of the ferret brain (see Fig. 2A).
After surgery, animals received saline (20 mL, sc), dexamethasone (0.5–1.0 mg/kg, im), and yohimbine (2 mg/kg, im). To avoid pain and infections, for 3 days postoperatively, ferrets were treated with buprenorphine (0.02 mg/kg, im) and enrofloxacin (2.5–5.0 mg/kg, im).
Choice of Tracer, Preparation, and Injections
The bulk of the study presented here is based on the use of 10 kDa (mostly an anterograde tracer) and 3 kDa (mostly a retrograde tracer) BDAs. WGA–HRP, FG, and FB were used for control experiments and confirmation of results obtained with BDA. WGA–HRP transports both retrogradely and anterogradely; FG and FB are primarily retrograde fluorescent tracers (Schmued and Heimer 1990; Reiner et al. 2000). It was empirically determined that the revelation of fine details by both BDA tracers and both fluorescent tracers required the animal to survive for approximately a week, whereas the best results for labeling of somata by WGA–HRP were obtained after 24 h. BDA and WGA–HRP were prepared in sterile injectable saline as 10% and 1.25% solutions, respectively. FG and FB were prepared in ddH2O as 4% and 2.5% solutions, respectively. A volume of 0.05–0.3 μL of BDA, FG, or FB was injected per site. Because in our experience WGA–HRP spreads more than other tracers, injections of this tracer were 0.03–0.12 μL (see Supplementary Table 1).
First, large (∼0.3 μL) injections of BDA were applied into MD in order to cover all or most of this nucleus, and to insure that the cortical projections to the frontal pole would cover all or most of what could then be considered to be PFC. Progressively smaller injections were then applied in an effort to distinguish between fibers coming from the medial or magnocellular MD and those coming from the lateral or parvocellular MD. Once the frontal cortical areas that received input from the MD were determined, small to midsize (∼0.05–0.15 μL) injections were placed in these cortical areas in order to determine if they projected back to the MD areas from which they received the bulk of their thalamic input.
Perfusion and Tissue Preparation
BDA and Fluorescent Tracers Injection Cases
Animals were left to survive 5–10 days, euthanized with an overdose of sodium pentobarbital (65 mg/kg, ip), and transcardially perfused with saline followed by cold 4% paraformaldehyde plus 0.1–0.2% glutaraldehyde in 0.1 M phosphate buffer (PB, pH 7.4). Brains were blocked and postfixed overnight in 4% paraformaldehyde followed by a series of 10%, 20%, and 30% sucrose for 24 h each. Coronal or sagittal sections 40–70 μm thick were cut in a freezing microtome and collected in 0.1 M PB.
WGA–HRP Injection Cases
After 24-h survival, animals were euthanized and transcardially perfused with saline followed by a cold solution of 1% paraformaldehyde and 1.25% glutaraldehyde in 0.1 M PB. The perfusion continued with cold sucrose solutions (5–10%, 200 mL each). The brain was removed from the skull, then blocked and left in 20% sucrose (2 h, 4 °C) followed by 30% sucrose (4 °C for 24–72 h). Afterward, the blocks were frozen and cut as described above. All sucrose solutions were prepared in 0.1 M PB.
Histology and Immunohistochemistry
Sections containing BDA were treated with 0.1 M glycine (5 min, RT, pH 7.4) followed by 0.5% hydrogen peroxide (5 min, RT), rinsed with 0.1 M PB, and incubated overnight in avidin–biotin peroxidase complex (ABC, 1:400, at 4 °C; ABC kit, Vector Laboratories, Inc., Burlingame, CA) with 0.5% Triton X. Sections were developed using 3,3′-diaminobenzidine tetrahydrochloride (DAB) intensified with nickel (Ni) and mounted on gelatin-coated glass slides. Sections containing FG and FB were coverslipped either with DPX or vectashield (Vector Laboratories, Inc., Burlingame, CA). Sections containing WGA–HRP were processed by the tetramethyl benzidine method (Mesulam 1978). In most cases one forth to one-half of the sections were counterstained with cresyl violet, neutral red, or thionin to aid in viewing cytoarchitectural details. Occasional staining for different calcium-binding proteins, peptides, and enzymes (e.g., acetylcholinesterase, parvalbumin, calbindin, calretinin, cytochrome oxidase, substance P, and other) were also performed mostly to aid in the recognition of cytoarchitectural details.
For every case, a representative sample of at least 25% of all sections was photographed. Hence, in cases in which sections were 70 μm thick the maximum distance between documented sections was 210 μm. Areas of high interest (i.e., injection sites) or areas with heavy anterograde innervations or retrogradely labeled neurons were sampled more often (from all to half of the sections were documented). Photomicrographs of whole sections were obtained using microcomputer imaging device (MCID) software (Imaging Research, Inc., Ontario, Canada) via a light box and a Dage MTI CCD72 camera and control system (DAGE-MTI of MC, Inc., Michigan City, IN). The brightness and contrast of photomicrographs were occasionally enhanced in Photoshop. Drawings were made in Adobe Illustrator and their accuracy was checked in a small sample of sections that were also mapped using the Neurolucida software program (MicroBrightField, Inc., Williston, VT) at ×40 magnification. For this, the Neurolucida software interfaced with a computerized Zeiss Axioplan microscope system (Carl Zeiss MicroImaging, Inc., Thornwood, NY).
Tracing and Analysis of the BDA Injections and Projections
BDA injection sites were easily visualized even without the help of a microscope due to the heavy precipitation of Ni-DAB. Optical density measurements were taken (using MCID software) to determine the volume of tissue covered by the injected tracer and its dorsoventral versus mediolateral spread. Calculation of total tissue volume impregnated was done based on at least half of the sections.
Areas that received a heavy density of projections were also easily visualized even without the help of a microscope. Target areas of high optical density contained either projection fibers or some retrogradely labeled cells. Where a single heavy density area contained both retrogradely labeled somata and anterogradely labeled axons, the difference was observed under the microscope and the drawing of the corresponding areas was done manually.
Sources of Artifact
Tracer injections were steady and slow (over 10–15 min) to minimize tissue damage and spread and the needle was left in place for at least 5 min after the end of the injection to minimize flow up the needle track. The syringe was mounted with a Hamilton Chaney adapter to improve stability. Although this technique helps to minimize tissue damage, some uptake of BDA by damaged axons (Brandt and Apkarian 1992) and intact axon terminals (Jiang et al. 1993) will occur. In practice, these possibilities appear to have been minimal in our hands and do not seem to prevent a clear interpretation of the results. It is known that 10-kDa BDA, in the way we prepared it, transports mainly anterogradely although it also yields some retrograde labeling (Veenman et al. 1992), whereas 3-kDa BDA transports mainly retrogradely with very little anterograde transport. Because we were interested in the reciprocal connections between MD and PFC, the presence of both retrograde and anterograde transport is an advantage.
Stereotaxic Coordinates of MD and PFC
In order to be able to consistently place tracer injections in MD and/or PFC a custom made atlas of the ferret brain that uses Bregma as the main landmark was used (Fig. 2A). The set of coordinates we use apply only to male ferrets 6–14 weeks of age, when Bregma is clearly visible. After approximately 14 weeks of age, the sagittal crest starts obscuring Bregma making it very difficult or impossible to use as a landmark. There is great gender dimorphism in this species and even considerable differences between individuals of the same sex, a fact that makes a generalized stereotaxic atlas very difficult to attain (Lawes and Andrews 1987).
Our account of the ferret MD territories is based on our own histological observations, comparisons made to the description of the ferret thalamus by Herbert (1963) and by homology to the cat's thalamus as described by Berman and Jones (1982). Projections from MD to PFC helped us elucidate approximate stereotaxic coordinates for the PFC whose caudal limit in the dorsal surface of the brain is approximately 10 mm rostral to Bregma at 1 mm lateral to the midline and which extends along the cortex surrounding the presylvian sulcus on the orbital gyrus. For convenience, Figure 1 provides an overall description of the main surface landmarks of the ferret brain as well as its relationship to the skull.
The MD and PFC injections of BDA resulted in strong labeling with a dense core of labeled neurons and neuropil at the center of the injection and a surrounding aura of gradually decreasing intensity. The most opaque injection core was assumed to be the zone of most effective tracer uptake and in all cases presented here the injection cores were always well within the bounds of the MD or cortical area targeted. This was also the case for other tracers.
Border between Magnocellular and Parvocellular MD
The extent and location of the ferret MD with respect to other thalamic structures (Figs 2, 3, and 9B2), and the difference in cellular composition within the MD (Fig. 3) that was observed in this study are in agreement with the description of ferret MD provided by Herbert (1963). In short, of the thalamic nuclei of the medial group, the MD is the largest and occupies most of the medial thalamus. Rostrocaudally MD extends from the level of the anterior nuclei to the level of the posterior thalamic nuclei. At about midway in its rostrocaudal extent it reaches its largest size. At this level, it is bound medially by the midline nuclei and laterally by the anterolateral nucleus and the posterior pole of the anteroventral nucleus. Ventrally it is bound by the intralaminar nuclei in the internal medullary lamina. At this level, the cellular difference between the lateral or parvocellular component and the medial or magnocellular component is more easily distinguishable. However, as was also noted by Herbert, occasionally large cells are seen in the lateral component. The difference between magnocellular and parvocellular moieties is most clear dorsally. Ventrally the cells appear more intermediate in size and more densely packed than the cells in the lateral moiety. At the rostral and caudal extremes of the nucleus it is more difficult to see differences in cellular composition or arrangement. In general, rostrally the neurons are smaller and less densely packed than the cells of the nearby anteromedial nucleus (see Fig. 3A). Myelin staining is light on the medial part and appears more patchy laterally (see Fig. 3B). Caudally the MD is traversed by several fiber tracts and appears as a mixture of neurons of different sizes. The darkly stained habenular nuclei and the habenulo-peduncular tract or Retroflex bundle, which traverses the MD more or less through the medial part of the nucleus, provide excellent landmarks for the distinction of MD at this level. The centromedial nucleus appears more intensely stained than the MD (see Fig. 3C,D). At this level, the borders with other ventral and lateral nuclei are difficult to discern clearly. The cellular composition within MD as well as between MD and other thalamic nuclei constitutes a continuum of cells in which cytoarchitectonic differences usually take place in a distributed manner instead of abruptly. To facilitate the viewing of differences in cytoarchitecture, a combination of different stainings is most useful. In this study, in addition to Nissl and myelin stainings (Fig. 3) we also used AChE (Fig. 2), PV (Fig. 9B2), SMI-32, Substance P (Fig. 5D), and other.
BDA injections into MD varied in size from approximately 0.5–7.5 mm3 of tissue. Out of 17 injections into MD, relative to each other, 6 injections were considered large, 6 were midsize, and 5 were rather small. Twenty-two BDA injections into the PFC were small to midsize (0.9–3.5 mm3). Here we describe specific cases that illustrate the results obtained from these 17 MD and 22 PFC injections (see Supplementary Table 1).
Tracer Injections into MD and their Corresponding Projections to PFC
For illustration we show 5 representative cases, organized from largest to smallest, in which a mixture of 10- and 3-kDa BDA was injected into the MD. Figures 4 and 5 illustrate the results of a large (7.6 mm3) BDA tracer injection into the MD. The bulk of the injection (green area) was in the medial and lateral moieties of the MD along the entire dorsoventral extent, although some of the tracer infiltrated the dorsal edge of the ventral nuclei. Examination of the frontal cortex in parasagittal sections revealed both areas of innervation (black and red) as well as retrogradely labeled cell bodies (blue) in the anterior sigmoid and orbital gyri (which are separated by the presylvian sulcus). Comparing the location of fibers and cell bodies to Nissl, SMI-32, and Substance P, stained adjacent sections in PFC confirmed that the fibers were most dense in layer IV, whereas retrogradely labeled cell bodies are located in layer VI (Fig. 5). The course of the fibers was along the walls of the presylvian sulcus in the rostral aspects of the anterior sigmoid gyrus, followed by extension into the orbital gyrus.
Figures 6–8 illustrate (coronally) the extent and analysis of additional large to small (7.0, 5.0, 3.33, and 0.98 mm3) BDA injections into MD. In the case illustrated in Figure 6, the injection covered the total dorsoventral extent of the MD caudally and most of the ventral MD and approximately three quarters of the dorsal MD rostrally (Fig. 6A). At approximately the middle of the injection's rostrocaudal extent the tracer infiltrated also parts of the centrolateral nucleus and the dorsalmost edge of the ventromedial nucleus. The case in Figure 7 leaked minimally into the ventral nuclei and was not found in the centrolateral nucleus at all.
The amount of PFC labeling correlates clearly with the size of the injection into the MD. Larger injections (Figs 4, 6, 7) gave rise to large numbers of fibers and cells labeled in different areas of the PFC. Smaller injections (Fig. 8) allowed the correlation between MD moieties and more restricted PFC areas. In general, the MD innervation of these cortical areas was most dense in layer IV. The overall appearance of the MD fibers in a sagittal (Fig. 4 and Supplementary Fig. 1) or coronal (Figs 6, 7 and 8) view is a discontinuous set of patches. However, superimposition of photomicrographs indicates that these afferent fibers into the PFC form a more or less continuous web with some small areas of the PFC more heavily innervated than others.
We did not find, in any of our MD injections, projections to the rostrolateral anterior sigmoid gyrus. Very rarely, if at all, were stained fibers found in the rostrobasal part of the cortex, medial to the rhinal fissure. Similarly, neither stained fibers nor neurons were observed medial to the olfactory sulcus in the mid and ventral gyrus rectus (see Fig. 11). The areas dorsal and lateral to the presylvian sulcus at the rostrocaudal level where the posterior sigmoid gyrus is already well developed (coronal sections at approximately 9 mm rostral to Bregma) do not appear to connect reciprocally with the MD in any significant way. The bulk of the MD-cortical fibers were therefore found rostrally in the orbital gyrus, from where they appear to progress rostrocaudally from lateral to dorsomedial areas (Figs 10 and 11).
In general, the retrogradely labeled neurons were more broadly distributed than the areas of anterograde innervation, although the regions showing the most intensely retrogradely labeled cell bodies often corresponded with the regions of intense axonal innervation (see Figs 1, 6, 7 and 8). Localized injections of BDA into restricted portions of the MD (Fig. 8) suggested that different parts of this thalamic nucleus project onto different PFC regions. Reciprocal connections were confirmed by the presence of retrogradely labeled cells in those regions as well as by tracer injections directly into those areas (Fig. 9).
Smaller and restricted injections were always confirmed in separate animals. Six injections intermediate in size were placed in the middle of the MD in a transition area between magnocellular and parvocellular moieties. These injections resulted in a pattern of innervation that was intermediate between those illustrated in Figure 8.
Tracer Injections into PFC and their Corresponding Projections to MD
Once the areas in the frontal pole that received MD projections were identified, we proceeded to make small to medium size tracer injections into these cortical areas (n = 22), to test the hypothesis that they in turn project back to the areas from which they receive thalamic afferents. In general, injections in the PFC infiltrated all cortical layers and covered a volume of tissue ranging from approximately 0.9–3.5 mm3. Injections into the orbital gyrus of the rostrolateral PFC resulted in anterograde labeling in the medial (magnocellular) component of the MD (Fig. 9A; n = 3). In contrast, BDA injections into a more caudal and dorsomedial component of the PFC resulted in anterograde labeling in the parvocellular portion of the MD (Fig. 9C; n = 5). Cortical tracer injections that were located intermediate between these 2 extremes resulted in a pattern of projections that fell in between the rostral and medial (magnocellular) MD and the more caudal and lateral (parvocellular) MD (e.g., Fig. 9B).
Injections into thalamic nuclei near, but not including, MD resulted in either no or very few fibers in the areas of the PFC that receive innervation from the MD (n = 8 injections in 4 ferrets). These injections were in portions of the anteroventral, anterodorsal, lateral anterior, ventromedial, ventral anterior, internal medullary lamina, and one injection that also leaked into the third ventricle infiltrated the paraventricular nuclei. Therefore, it was concluded that areas around the perimeter of the MD nucleus of the ferret thalamus do not project in any significant way to the PFC areas as described here. A possible exception was indicated by a control injection into mostly the ventromedial nucleus (but which also leaked into the MD) which did produce a bit more fibers into PFC proper but not as many when compared with injections just into MD.
Control injections (6 injections in 2 ferrets) were also applied to cortical areas in the frontal pole other than the PFC. Here also it was found that despite intracortical and subcortical projections to different structures including the thalamus, there were no major reciprocal projections to MD. The occasional fibers encountered in MD and whose origins were outside the PFC were very sparse and could only be seen under high power microscopy. Agglomeration of afferent fibers or retrogradely labeled neurons was never intense enough to be seen with the naked eye, as was the case with injections applied into the PFC.
The goal of the present study was to determine the location and extent of the ferret PFC based on its heavy reciprocal connections with MD. Our findings indicate that the ferret PFC spans the dorsal and ventral walls around the presylvian sulcus rostral to the rostral tip of the anterior sigmoid gyrus and the dorsal and medial cortex at the level of the rostral anterior sigmoid gyrus. PFC is therefore located in the rostrodorsal anterior sigmoid gyrus and throughout the orbital gyrus (Figs 10, 11). Furthermore, our results indicate that the parvocellular or lateral component of MD projects to the dorsomedial component of the PFC, whereas the medial or magnocellular portion of MD projects to the more lateral component of the PFC. These portions of the PFC in turn project back to the areas in MD from which they receive the bulk of their afferents. To our knowledge this is the first report on the location and extent of the ferret PFC.
The Definition of PFC
Guidelines used in the past to aid in the characterization of PFC include the distinction of a clearly granular layer IV, and heavy innervation by thalamic MD fibers and dopamine afferent fibers from brainstem nuclei. However, these criteria and even the MD innervation itself, used here to define PFC, have previously been seriously called into question (Preuss 1995).
Because dopamine innervation was convergent with MD innervation in rat and other species it was suggested that dopamine innervation by itself may be sufficient criteria to characterize the PFC in the mammalian brain (Divac, Bjorklund, et al. 1978). However, it was later recognized that dopamine innervation in the frontal lobe of primates was widely distributed and that some areas in the dorsolateral PFC, including the principal sulcal cortex, were rather weakly innervated as compared with motor areas (Gaspar et al. 1992; Williams and Goldman-Rakic 1993; Preuss 1995). In terms of MD innervation and the presence of a granular layer, the classical primate PFC does receive heavy innervation from MD but only the dorsolateral PFC has a clear granular layer (whereas the orbitofrontal does not). In rat, the PFC is rather homologous to the premotor, anterior cingulated, and orbital cortex of macaques (Preuss 1995). Therefore, rats do not have a dorsolateral PFC per se, although they do have an agranular cortical region that some authors recognize as PFC. Although the classical PFC corresponds to what Brodmann (1909) clearly distinguished as granular cortex, later work giving more weight to the MD connections and less importance to the distinction of a granular layer included agranular cortical areas as PFC (Walker 1940). Architectonically, the region considered here to correspond to the PFC of the ferret seems similar to that of many other species, apart from the dorsolateral PFC of primates, in the sense that it does not appear to have a well-developed granular layer (see Fig. 5). In fact, comparison in the same parasagittal section between granular layer IV in the visual cortex of the ferret and areas in the orbital gyrus and the walls surrounding the presylvian sulcus makes it evident that the granulation in layer IV of the ferret PFC is rather poor. However, small granular cells are present as observed in Nissl and Golgi (not shown) staining, and comparison between these and bigger pyramidal cells present in adjacent layers III and V makes it possible to distinguish the location of layer IV (see Fig. 5).
It is now recognized that the MD does not only send projections to the PFC but also to many other cortical and subcortical structures (Kievit and Kuypers 1977; Akert et al. 1979; Goldman-Rakic and Porrino 1985; Vogt et al. 1987; Giguere and Goldman-Rakic 1988; Matelli and Luppino 1996). In the absence of more reliable guidelines it is our opinion that the heavy reciprocal connections between MD and areas of the frontal pole still constitute the best anatomical approach to characterize the extent of the PFC. Historically, either afferent and/or efferent connections with/to the MD have been used to study the PFC in several species including humans (Meyer et al. 1947; McLardy 1950), nonhuman primates (Walker 1936; Akert 1964; Tobias 1975; Kievit and Kuypers 1977), opossum (Divac, Bjorklund, et al. 1978; Benjamin and Golden 1985), sheep (Rose and Woolsey 1948), rabbit (Rose and Woolsey 1948; Benjamin et al. 1978), dog (Akert 1964; Narkiewicz and Brutkowski 1967), cat (Rose and Woolsey 1948; Markowitsch et al. 1978, 1980), guinea pig (Markowitsch and Pritzel 1981), rat (Leonard 1969; Krettek and Price 1977; Divac, Kosmal, et al. 1978; Conde et al. 1990), and mouse (Guldin et al. 1981).
Location of the PFC in the Ferret
The ferret PFC lies on the rostral part of the brain, within the rostralmost part of the anterior sigmoid gyrus and the orbital gyrus (Figs 10, 11). Most of the anterior sigmoid gyrus in ferret is presumed to be motor cortex (Lockard 1985) and at least some of the fibers that innervate the most caudal aspect of the anterior sigmoid gyrus may come from the ventral thalamic nuclei that are known to innervate the motor and supplementary motor cortices in other species. On the medial surface, on a midsagittal plane, the most caudal part of the PFC appears in the most rostral part of the anterior sigmoid gyrus immediately posterior to where the presylvian sulcus would meet the longitudinal fissure if, at this level, the presylvian sulcus would go all the way to the midsagittal plane (this is approximately 10–11 mm anterior to Bregma for male ferrets 6–14 weeks of age). This band of fibers, close to the medial wall, extends rostrally to immediately before the rostralmost tip of the brain and then turns ventrally into what becomes the olfactory sulcus (see L 1.54 Fig. 4 and section nos. 1–3, Fig. 7). It spans the dorsal aspect of the brain close to the medial wall at the level where the orbital gyrus and the corresponding part of the dorsal and medial wall of the rostral tip of the anterior sigmoid gyrus meet (Fig. 11). The PFC then extends laterally into the dorsal and ventral walls of the presylvian sulcus (see L 3.36 Fig. 4 and Figs 10, 11), and spans also the fundus of this sulcus. The fibers disappear from the dorsal surface of the anterior sigmoid gyrus and reappear again dorsally at more lateral levels. Some fibers lie in the rostral and ventral aspect of the orbital gyrus. Therefore, because the orbital gyrus and particularly the walls around the presylvian sulcus appear to possess the heaviest reciprocal connections with the MD they constitute most of what can be considered PFC in the ferret. However, one should keep in mind that our study was done in young male ferrets and future studies are necessary to confirm their validity in fully developed adult ferrets of both sexes.
Phylogeny and Comparative Anatomy between the Ferret PFC and that of Other Species
Because the definition of the PFC relies strongly on its connections with the MD nucleus of the thalamus, MD similarities and differences between species could then reflect corresponding similarities and differences in PFC, particularly in regards to connections between different PFC areas and the parvocellular and magnocellular moieties of the MD. Below, we describe first some of the interspecies similarities and differences with regards to the MD and its prefrontocortical connections and then the interspecies similarities and differences of the PFC per se. We are aware of current controversy regarding phylogeny and, particularly, its relation to the question of the function of the PFC (Wise 2008), a full discussion of which is beyond the scope of this paper.
Mostly based on Nissl and Golgi stainings alone or in addition to retrogradely labeled neurons and anterogradely labeled fibers, most investigators recognize at least some separate and distinct cytoarchitectonic mediolateral, dorsoventral, and rostrocaudal components of MD in different species. With regard to the primate thalamus, Olszewski (1952) gave one of the earlier descriptions of different MD moieties, although, earlier magnocellular and parvocellular components and their differential projections to the PFC had been recognized (Clark 1930; Walker 1938). The parvocellular part of MD was found to project to the frontal granular cortex in the dorsolateral surface, whereas the medial magnocellular part projects mainly to the orbitofrontal cortex of the medial surface of the frontal pole (Walker 1938). The topographical organization of MD-cortical connections is therefore similar between ferrets and primates, although, in ferrets the dorsal PFC appears relatively medial in comparison to the primate dorsal PFC.
In cat, the MD has been divided into dorsal, ventral, medial, and lateral components (Niimi and Kuwahara 1973). In the rat, the MD was divided into central, medial, and lateral components (Krettek and Price 1977) within which some authors recognized further subdivisions such as intermediate, lateral (Velayos and Reinoso-Suarez 1985), and even an extreme lateral paralammelar one (Leonard 1972; Groenewegen 1988). In general, information on the homology between the different moieties of the rodent or feline MD and those of the primate is very limited (Kuroda et al. 1998). It seems to be the case that, at least tentatively, the most medial and central parts of the rat or cat MD correspond to the magnocellular division of the primate MD, whereas the more lateral components of the rat or cat MD correspond to the parvocellular division of the primate MD (Kuroda et al. 1998); a relationship that appears to hold true for other species, including ferrets. Of particular interest for our studies in the ferret is the fact that the dorsal MD of the cat projects to the medial PFC, whereas the more ventral portion of MD projects to the lateral PFC (Khokhryakova 1978; Niimi et al. 1981). This is similar to the results we obtained here. We observed that the medial part of the ferret MD lies more ventral, whereas the lateral MD appears more dorsal. Hence, the medial MD of the ferret appears to be homologous to the cat's MD ventral division. This medial and more ventral division of the ferret MD projects to and receives projections from the more rostral and lateral aspect of the PFC (see Figs 8A and 9A), whereas the more lateral and dorsal division of the ferret MD, which probably corresponds to the cat's dorsal division, projects to and receives projections from the more caudal and dorsomedial component of the ferret PFC (see Figs 8C and 9C).
The very clear connection we describe between the MD and the orbital gyrus of the ferret is also present in dogs (Narkiewicz and Brutkowski 1967; Kosmal and Dabrowska 1980), rats (Divac, Kosmal, et al. 1978), rabbits (Benjamin et al. 1978), and primates (Goldman-Rakic and Porrino 1985). Although, such a connection is lacking in cats, this apparent discrepancy, as noted by Kreiner (1970) and later confirmed by Markowitsch et al. (1980), may simply be due to a relative topographical difference between the presylvian sulcus and the orbital gyrus. In cat, the orbital gyrus is dorsal to the presylvian sulcus (Kreiner 1970), whereas in other species (dog, ferret) the orbital gyrus is ventral to the presylvian sulcus. The orbital gyrus of the cat does not receive projections from the MD. However, as Markowitsch et al. (1980) noted, if only the presylvian sulcus is taken as the topographical point of reference in both cats and dogs, then the mediodorsal cortical afferents are indeed similar between these and other species because the cortical area ventral to the presylvian sulcus is indeed innervated by the MD.
Although not always accepted, it has been (and continuous to be) argued that PFC extent increases with phylogenetic development (i.e., Broadman 1912; Blinkov and Glazer 1968; Fuster 1980; Semendeferi et al. 2001, 2002). In human, more than in any other animal, the PFC occupies a greater proportion of the cerebral cortex, a fact interpreted as suggestive of its role in higher cognitive functions that distinguish humans from other animals (Ariens Kappers et al. 1960; Radinsky 1969; Fuster 1980). Within a phylogenic order, cortical evolution can be studied by homologies between sulci and gyri (Fuster 1980), which also serve as a basis for interspecies comparisons. When comparing different members of the carnivora family, the presylvian sulcus is one of the most consistent sulci (Ariens Kappers et al. 1960) and it is homologous to the vertical limb of the arcuate sulcus of monkeys and the inferior precentral fissure of humans (Fuster 1980). The sulcus principalis of monkeys or sulcus frontomarginalis of man appears to be homologous to the proreal or intraproreal fissure of some carnivora (Ariens Kappers et al. 1960; Fuster 1980). To our knowledge, in ferret, a proreal or intraproreal sulcus has not been identified. However, there is a deep olfactory sulcus that in the rostroventral part of the ferret brain separates the orbital gyrus from the gyrus rectus (Nigel et al. 1998).
The cat PFC has been studied in some detail, including its anatomical subdivisions (Markowitsch et al. 1978), structural organization (Khokhryakova 1978; Babmindra et al. 1979), and extent (Markowitsch et al. 1980). It has also been compared with that of primates (Khokhryakova 1978). Laterally, and similarly to the ferret, the PFC of the cat occupies most of the rostrocaudal extent of the fundus and medial wall of the presylvian sulcus and the proreal gyrus while medially, it occupies most of the surface of the frontal and rectus gyri (Cavada and Reinoso-Suarez 1985). According to our results the ferret PFC involves the orbital gyrus, which has substantial reciprocal connections with the MD. However, none of the injections placed in the MD labeled fibers or cells in the gyrus rectus and cortical injections were not placed in the gyrus rectus. Therefore, this area requires further studies to elucidate its cytoarchitectonics, connectivity, and function. Also, although in ferret our results indicate that the banks of the presylvian sulcus are part of the PFC, in dog, Tanaka (1987) points out that according to Kosmal et al. (1984) the banks of the presylvian sulcus are thought of as a transition area between PFC and motor cortex, at least in terms of cytoarchitecture and corticocortical connections.
Because, in gross anatomical terms, the PFC of the monkey is divided in at least 3 areas associated with distinct behaviors (the dorsolateral with spatial and mnemonic functions, the orbitofrontal with motivational and emotional functions and the frontal eye fields with visual attention and saccadic movements; Teuber 1972), future investigations of the ferret PFC should consider, among other things, these important anatomical–functional relationships.
National Institutes of Health.
This work was inspired and supported by Dr Patricia S. Goldman-Rakic (deceased), to whom we are forever grateful. We thank Ms Miriamma Pappy for help in the preparation and processing of tissue and for her excellent technical assistance. We thank a handful of individuals who contributed their comments to previous versions of this manuscript.
Conflict of Interest: None declared.