Clinical manifestations in diseases affecting the dopamine system include deficits in emotional, cognitive, and motor function. Although the parallel organization of specific corticostriatal pathways is well documented, mechanisms by which dopamine might integrate information across different cortical/basal ganglia circuits are less well understood. We analyzed a collection of retrograde and anterograde tracing studies to understand how the striatonigrostriatal (SNS) subcircuit directs information flow between ventromedial (limbic), central (associative), and dorsolateral (motor) striatal regions. When viewed as a whole, the ventromedial striatum projects to a wide range of the dopamine cells and receives a relatively small dopamine input. In contrast, the dorsolateral striatum (DLS) receives input from a broad expanse of dopamine cells and has a confined input to the substantia nigra (SN). The central striatum (CS) receives input from and projects to a relatively wide range of the SN. The SNS projection from each striatal region contains three substantia nigra components: a dorsal group of nigrostriatal projecting cells, a central region containing both nigrostriatal projecting cells and its reciprocal striatonigral terminal fields, and a ventral region that receives a specific striatonigral projection but does not contain its reciprocal nigrostriatal projection. Examination of results from multiple tracing experiments simultaneously demonstrates an interface between different striatal regions via the midbrain dopamine cells that forms an ascending spiral between regions. The shell influences the core, the core influences the central striatum, and the central striatum influences the dorsolateral striatum. This anatomical arrangement creates a hierarchy of information flow and provides an anatomical basis for the limbic/cognitive/motor interface via the ventral midbrain.
The nucleus accumbens plays a major role in mediating motivation and reward. Studies of this striatal region have focused on its role in influencing motor outcome by funneling information from the limbic system to the motor system (the “limbic/motor interface”) (Nauta and Domesick, 1978; Mogenson et al., 1980; Heimer et al., 1982; Kalivas et al., 1993). Nauta first proposed that dopamine plays a role in this limbic/motor interaction through the accumbal projection to the substantia nigra, which in turn projects to the dorsal striatum (Nauta and Domesick, 1978; Nauta et al., 1978; Somogyi et al., 1981; Haber and Fudge, 1997). However, the dorsal striatum is involved in more than motor function. In primates it is linked not only to motor and premotor cortical areas but to all of frontal cortex, including the dorsolateral prefrontal cortex.
Motor and premotor cortex projects only to a small dorsolateral striatal region at rostral levels and to much but not all of the putamen centrally and caudally (Künzle, 1975, 1978; Flaherty and Graybiel, 1994). Most of the head of the caudate nucleus and the rostral putamen receives input from the dorsolateral prefrontal cortex, which is involved in working memory (Goldman-Rakic and Selemon, 1986;Goldman-Rakic, 1994). The ventromedial striatum (VMS), which includes the nucleus accumbens, and the rostral, ventral caudate nucleus, and putamen, receives its frontal input from the orbital and medial prefrontal cortex (OMPFC) (Kunishio and Haber, 1994; Haber et al., 1995a; Chikama et al., 1997). The OMPFC (including the anterior cingulate cortex, medial and lateral orbital cortex, and agranular insular cortex) plays a key role in the development of reward-guided behaviors by linking primary rewards with motivation and emotion (Rolls et al., 1980; Eslinger and Damasio, 1985; Fuster, 1989; Carmichael and Price, 1994; Carmichael and Price, 1996). As in rodents, the ventromedial striatum in primates contains two subdivisions: the “shell,” distinguished by its calbindin-negative staining and limited input from the cortex, midbrain, and thalamus, and the “core,” which is histochemically indistinguishable from the rest of the striatum (Zaborszky et al., 1985; Zahm and Brog, 1992; Kunishio and Haber, 1994; Lynd-Balta and Haber, 1994b; Giménez-Amaya et al., 1995; Haber et al., 1995a; Groenewegen et al., 1996; Meredith et al., 1996; Voorn et al., 1996; Chikama et al., 1997; Heimer et al., 1997). Thus, the frontostriatal projection pattern is organized in a ventromedial to dorsolateral gradient, from limbic and cognitive to motor functions. Given this frontostriatal organization, we sought to reexamine the limbic/motor interface via the substantia nigra (SN) neurons to determine how information across limbic, cognitive, and motor circuits is integrated via the striatonigrostriatal (SNS) pathways.
Dopamine neurons, which comprise the majority of SN pars compacta cells, are considered to be key for focusing attention on significant and rewarding stimuli, a requirement for the acquisition of new learned behaviors (Grace and Bunney, 1995; Schultz et al., 1997; Yamaguchi and Kobayashi, 1998). This acquisition not only involves limbic, cognitive, and motor striatal pathways, it requires the coordination of these functions. For a behavioral response to occur to a particular stimulus, information about motivation and reward as well as cognition are required to execute the appropriate movement. Studies of integration between these circuits have focused on how the limbic system directly modulates motor outcome (Mogenson et al., 1980; Jimenez-Castellanos and Graybiel, 1989; Kalivas et al., 1993; Mogenson et al., 1993; Gerfen and Wilson, 1996; Groenewegen et al., 1996; Haber and Fudge, 1997). However, a direct limbic–motor interface does not consider the entire striatal system, including the cognitive component. We studied the organization of SNS pathways as they relate to the OMPFC, dorsolateral prefrontal, and motor corticostriatal input. Furthermore, previous studies have shown the organization of either the nigrostriatal pathways or the striatonigral pathways. Our goal was to determine how the entire SNS subcircuit directs information flow through the shell, ventromedial, central, and dorsolateral striatal regions.
MATERIALS AND METHODS
There were two sets of experiments. The first set placed bi-directional tracers into different regions of the striatum (see Fig.1 a). These cases were analyzed first for the distribution of labeled cells in the frontal cortex. On the basis of the cortical labeling pattern, each injection site was classified as follows. (1) “Motor” striatum were injection sites that labeled cells primarily in frontal cortical areas 4 and 6 with few labeled cells in areas 9 and 46, and scattered cells, or none, in orbitofrontal regions or in areas 32, 25, 24, a or b; (2) “limbic” striatum were injection sites that labeled cells primarily in areas 32, 25, 24, a and b, and medial orbitofrontal cortex, areas 10, 14, and 13, with few labeled cells in areas 9 and 46 and none in areas 4 and 6. We defined the shell as the ventral striatal region that was calbindin (CaBP) negative and the rest of the ventromedial striatum as the “core” (Meredith et al., 1996). (3) Association areas comprised injection sites that labeled primarily areas 9 and 46. Each case was classified according to its cortical labeling pattern, and anterograde tracers were then placed into the striatum to match the retrograde placements in motor, limbic, or association areas. These cases were charted for cell and fiber labeling pattern in the midbrain. In some cases two tracers were injected into the same or different striatal regions in the same animal.
The second set of experiments placed anterograde and bi-directional tracers in different regions of the ventral midbrain (see Fig.1 b). The midbrain dopamine cells are divided into a dorsal and ventral tier. The dorsal tier, including the dorsal pars compacta and the ventral tegmental area (VTA), is calbindin positive. The ventral tier includes the densocellular region and the cell columns and is calbindin negative (Fallon et al., 1978; Gerfen et al., 1987; Lavoie and Parent, 1991; Lynd-Balta and Haber, 1994a; Haber et al., 1995b). Sections containing injection sites were counterstained for calbindin to determine their location.
Animals and procedures. Adult Macaques (Mulataand Nemistrina) were injected with one or more tracer molecules: anterograde tracers, Phaseolus vulgarisleucoagglutinin (PHA-L) and tritiated amino acids, and bi-directional tracers, wheat germ agglutinin conjugated to horseradish peroxidase (WGA/HRP), Lucifer yellow (LY), or fluorescein (FS) conjugated to dextran amine. After initial anesthesia with ketamine (10 mg/kg, i.m.), a deep surgical level of anesthesia was maintained with pentobarbital (initial dose 20 mg/kg, i.v., and maintained as needed). Targets were located using electrophysiological mapping. Serial electrode penetrations were made throughout the rostrocaudal and mediolateral extent of the striatum to identify neuronal activity based on patterns of electrophysiological recordings (Haber et al., 1993). The location of neurons encountered in a series of penetrations was used to prepare a map indicating the boundaries of different basal ganglia structures. The absence of cellular activity was noted in the area of fiber tracts, i.e., the corpus callosum, the internal capsule, and the anterior commissure. Accurate placement of the tracers was subsequently achieved by careful alignment of the injection cannulae with the electrode. Tritiated amino acids (tritiated leucine and tritiated proline, 50–80 mCi, in 200 nl saline; NEN, Boston, MA.), PHA-L, 80 nl of 2.5% in 0.05m Tris buffer (Vector Laboratories, Burlingame, CA.), LY, 20–40nl, FS, 40–50 nl, (10% in dH2O; Molecular Probes, Eugene, OR.), and HRP-WGA, 40–50 nl, (4% in H2O, Sigma, St. Louis, MO.) were pressure-injected into discrete regions of the striatum or midbrain. After an injection, the syringe remained in situ for 20–30 min to prevent leakage up the needle track. Nine to 14 days after surgery, the animals were again deeply anesthetized and perfused through the heart with saline followed by a 4% paraformaldehyde/1.5% sucrose solution in 0.1 m phosphate buffer, pH 7.4. The brains were cryoprotected in increasing gradients of sucrose (10, 20, and finally 30%). Serial sections of 50 μm were cut on a freezing microtome and processed for autoradiography or immunocytochemistry for WGA-HRP, PHA-L, LY, or FS. Sections were also double-labeled for two tracers (see below).
Sections for autoradiography were mounted on chrome-alum gelatin-coated slides and defatted in xylene overnight. Slides were dipped in Kodak NTB2 photographic emulsion and exposed for 4–6 months at −20°C in a light-tight box. The sections were then developed in Kodak D19 for 2.5 min., fixed, washed, and counterstained with cresyl violet. Sections to be immunoreacted with anti-PHA-L, anti-LY, anti-FS, or anti-HRP-WGA were rinsed in 0.1 m phosphate buffer, pH 7.4, with 0.1%m Triton X-100 (PBS-T), preincubated in 10% normal goat serum (NGS) diluted with PBS-T for 30 min, and then placed in the primary antisera, anti-LY or FS (1:1000, Molecular Probes), or anti-HRP-WGA (1:2000; Dako, Carpinteria, CA.), or anti-PHA-L (1:500; EY Labs, San Mateo, CA.) in NGS-PBS-T for four to five nights at 4°C. The avidin–biotin reaction (rabbit Vectastain ABC kit, Vector) was used to visualize the tracers. Staining was produced by incubating the tissue for 10–12 min in 3,3′ diaminobenzidine tetra-hydrochloride and 3% hydrogen peroxide and intensified with 1% cobalt chloride and 1% nickel ammonium sulfate to yield a black reaction product. Sections were rinsed, dehydrated, and coverslipped with Permount (Fisher Scientific, Pittsburgh, PA). Antisera to calbindin (Sigma) was used at 1:10,000 in PBS-T with 0.5% bovine serum albumin (BSA) (Sigma). Tissue was first incubated in PBS-T with 5% BSA for 1 hr, then incubated in primary antisera for four nights at 4°C and processed using the avidin–biotin reaction (mouse Vectastain ABC kit, Vector) as described above.
Analysis. Cases were eliminated from analysis if there was contamination of adjacent structures, including fiber tracks. Cell and fiber distributions in both the striatum and midbrain were charted for each case. To determine the scope of SNS interactions, we used different combinations of cases and analyzed experiments both individually and collectively in these groups. To create composites from all injection sites within a functional region, Nissl-stained midbrain or striatal sections were imported into the computer using NIH Image Software and a Hamamatsu camera (magnification 6.3×). Photomontage images were imported into the graphics program Canvas 5.0. Within this program, the following layers were created: layer 1, the photomontage of the Nissl-stained section; layer 2, a drawn outline of the section with internal landmarks such as the pars compacta, fascicles of the third nerve, the basis peduncle, the red nucleus, and the aqueduct; and layer 3, charts of the individual cells or fiber distributions. To evaluate the collective pathways from each striatal or midbrain region, a master chart was created for each rostral/caudal level charted. The master charts contained each individual case within its own layer. One Nissl photomontage for each level was imported into layer 1. The outline of each individual case along with its charted cell or fiber distributions was imported as a separate layer and superimposed on the master Nissl image. The outline for each case was aligned to best correspond to the outline of the master image. Once each case was imported, the individual outlines were eliminated, leaving the fiber and cell labeling for each case in its own layer. The relationship between the different input and output pathways across cases was analyzed by changing the visible layers. This allowed us to evaluate within and between each SNS system by combining data from discrete injections from different animals. Relationships between collective anterograde and retrograde injection sites were also compared with individual cases that contained bi-directional tracers in the relevant striatal regions.
Striatal cases were analyzed in which anterograde or retrograde tracer injections were placed into different striatal regions associated with limbic, cognitive, or motor corticostriatal pathways (Fig. 1 a). There were 16 bi-directional and anterograde injection sites into different parts of the midbrain, including the dorsal tier [both the VTA and the dorsal SN pars compacta (SNc)] and the ventral tier (both the densocellular region and the cell columns) (Fig. 1 b). Different tracer molecules, when placed in a similar position, resulted in a similar projection pattern. Thus, there were no differences between retrograde labeling patterns of HRP-WGA and LY; however, in general fewer cells were labeled using LY. Tritiated amino acids and LY injections resulted in denser terminal and fiber labeling than PHA-L injections, but with the same distribution pattern. There were no differences in labeling patterns or density between the two species of monkeys.
Taken together, there was a general inverse dorsal–ventral topography in both the striatonigral projections and nigrostriatal projections. Dorsal striatal injections resulted in labeled fibers and/or cells in the midbrain ventral to those labeled after a more ventral injection of tracer. Likewise, dorsal midbrain injections resulted in labeled fibers or cells in the ventral striatum. These general results are consistent with those reported previously (Szabo, 1962; Parent et al., 1984;Hedreen and DeLong, 1991; Lynd-Balta and Haber, 1994a,c; Parent and Hazrati, 1994). In this paper, we specify the characteristics and relationships between both limbs (the nigrostriatal and striatonigral) of the SNS system with particular reference to cortical innervation of the striatum.
Distribution of labeled cells in cortex
Injection sites were first classified according to cortical labeling patterns (Table 1). Injections placed rostral to the anterior commissure into the nucleus accumbens, the ventromedial caudate nucleus, and ventral putamen labeled cortical cells in medial areas 32, 25, 24, a/b, and regions of orbitofrontal cortex (areas 14, 13a/b, 13, and 12). We referred to the region of these sites as the VMS. Within this group, three sites were confined to the shell (CaBP-poor region), without contamination of surrounding areas, and three were placed in the shell but not confined to it. The injection sites placed into the dorsal shell (or cone) labeled areas 25, Ia, and 14. A more ventral injection site also labeled these areas and areas 32, 24a/b, and 13a/b. Injections into the medial ventral caudate nucleus labeled additional areas 11 and 12. Central core injection sites labeled many cells in lateral regions, including area 13 and fewer in medial areas 25, Ia, and 14. Injection sites placed into the head and central body of the caudate nucleus and into the central, rostral putamen resulted in labeled cells concentrated in the dorsolateral prefrontal cortex (areas 9 and 46). Those that included the ventral part of the central striatum (CS) labeled cells not only in the dorsolateral prefrontal cortex but also in lateral parts of the orbital cortex. In contrast, sites placed more dorsally in the central striatum labeled some cells in area 8 and premotor areas in addition to dorsolateral prefrontal cortex. The region of injections sites that labeled primarily dorsolateral prefrontal cortex were referred to as the CS. Retrograde tracers placed in the dorsolateral striatum (DLS) rostrally and throughout most of the central and caudal putamen resulted in labeled cells in motor, premotor, and supplementary motor (SMA) cortex, and the cingulate motor area (areas 4, 6, and 24c). We referred to the region of these sites as the DLS. One injection site, placed in the head of the caudate but in its dorsolateral corner, labeled cells specifically in the pre-SMA, which borders dorsolateral prefrontal cortex. The more dorsal putamen injection sites labeled cells predominately in areas 6m and 6d; ventral putamen injection sites labeled cells primarily in area 6v.
The VMS is that part of the striatum that receives input from the OMPFC, no input from motor or premotor areas, and little or no input from the dorsolateral prefrontal cortex. Projections from the VMS to the midbrain terminated in the dorsal midbrain, including both the dorsal tier and the dorsal part of the ventral tier extending into the medial and dorsal pars reticulata. Terminal fields compiled from all of the ventral striatal injection sites projected widely throughout the midbrain and were concentrated in the medial part rostrally and dorsolaterally at central and caudal levels (Fig.2). This general widespread terminal field is also evident after single injections (Fig.3 c). The midbrain cells that projected to the VMS were concentrated in the medial half of the midbrain.
Terminal labeling from the collective VMS injection sites overlapped extensively with the labeled cells that projected back to the ventral striatum (Fig. 2). However, the distribution of midbrain cells that projected to the VMS were more confined than the distribution of VMS projections to the midbrain, as evidenced also in individual cases. There were some labeled cells dorsal to the VMS terminal field. In addition, there was a large ventral terminal region, which did not contain labeled cells, that projected back to the VMS. Thus, there were three components in the SN of this projection system: Labeled cells that projected to the VMS that lie dorsal to the VMS efferent projection field (Fig. 2, black arrowheads); labeled cells that projected to the VMS that lie within the terminals fields of the VMS projection (Fig. 2, white arrowheads); and efferent VMS fibers that lie ventral to labeled cells projecting to the VMS (Fig. 2,arrows).
The shell SNS projection system lay in the most dorsal and medial part of the midbrain (Fig. 3). One interesting feature of the shell projection was that the fiber and cell labeling after injections centered in the shell did not obey the inverse dorsal–ventral topography. Labeled terminals and cells were always located in the dorsal SNc and in the VTA, regardless of the position of the injection site. Projections to the dorsal shell (or cone region) from the midbrain were concentrated near the midline, in the VTA (Fig.3 b), whereas the ventral shell injection site resulted in labeled cells more ventrolaterally. Projections from the shell to the midbrain were also concentrated in the VTA and dorsal tier. However, there were also some terminals among and surrounding the dorsal part of the densocellular region (Fig. 3 c, arrows). These terminals lay ventral to the cells that projected back to the shell. Projections to the core originated from cells in both the dorsal tier and the medial and dorsal densocellular region. Although this distribution of labeled cells overlapped with those cells projecting to the shell, it was somewhat ventral and more lateral to those cells projecting to the shell (Fig. 3 d). Terminal fields from the core lay within and ventral to the cells that projected back to the core (Fig. 4).
Both the shell and core SNS projection systems contained the three components in the SN (Figs. 3, 4): a dorsal group of labeled cells that projected to the shell or core but did not lie within its reciprocal efferent projection (Figs. 3, 4, black arrowheads); a group of cells that did lie within its efferent terminal field (Figs. 3, 4,white arrowheads); and a terminal field that did not contain reciprocally connected labeled cells (Figs. 3, 4, arrows). Efferent fibers from the shell terminated both within the region of labeled cells projecting back to the shell and ventral to it (Fig.3 a,c). Cells projecting to the core were located both within the terminal fields of the core and dorsal to it (Fig. 4). This dorsal group of cells lay within the ventral terminal field of the shell (Fig. 5 a,b). In contrast, there was little overlap between efferent fibers originating from the core and labeled cells projecting to the shell (Fig. 5 c). Labeled fibers from the core were also located ventral to the labeled cell population that projected back to the core.
The CS is the striatal region that receives input primarily from areas 46 and 9. However, in all cases some labeled cells were found in either the OMPFC or in area 6. The central SNS field lay primarily within the densocellular region of the midbrain (Fig.6 a–d). The CS efferent projection was extensive and terminated in the ventral part of the densocellular region and extended into the cell columns and surrounding pars reticulata. Labeled cells after retrograde tracer injections were also located throughout the densocellular region, in a wide medial–lateral area. As seen with the VMS, the collective CS SNS projection system also exhibited three components, a reciprocal and two nonreciprocal components. There were labeled cells located dorsal to the main CS terminal projection field (Fig.6 a, black arrowheads, and b). Labeled cells were found embedded within the CS terminal fields (Fig.6 a, white arrowheads, and c). Finally, terminals from the CS were found ventrally and not in close approximation to the main population of labeled cells that projected back to the CS (Fig. 6 a, arrows, andd). Cells that projected to the CS were generally located lateral and ventral to those labeled after VMS injection sites with some overlap. A few labeled cells extended into the dorsal cell columns.
The DLS is the region that receives input from areas 4, 6, and 24c, but not from the OMPFC, and little if any from areas 9 and 46. Efferent projections from the dorsal striatum were concentrated in the ventral and lateral half of the SN. Unlike the widespread terminal fields of the VMS and CS pathways, the distribution of labeled efferent fibers from the DLS pathway was more restricted. In contrast to its limited efferent projection, the distribution of labeled cells after retrograde injections into the dorsal striatum were widely distributed both in the cell columns and in the densocellular area. Similar to the organization of the other SNS systems, there were three components in the midbrain (Fig. 6 e): a dorsal group of cells that projected to the DLS but did not lie within its efferent projection field (Fig. 6 e, black arrowheads, andf); a group of labeled cells that did lie within its efferent projection (Fig. 6 e, unfilled arrowheads, and g); and a ventral terminal field that did not contain cells that projected back to the DLS (Fig.6 e, arrows, and h). The labeled cells in the densocellular region comprise the group that did not lie with the efferent projection from the DLS. This dorsal population of cells was relatively large, compared with that after VMS injections. Together, the DLS SNS pathway was made up of a widespread cell population that projected to the DLS and a relatively confined DLS efferent terminal field.
Relationships between VMS, DLS, and CS pathways
Projection fields from the VMS and DLS did not overlap in the midbrain (Fig. 7 a). Projections from the midbrain to the VMS arose from the dorsal tier cells, whereas the cell columns projected to the DLS. The densocellular group contained cells projecting to either or both striatal regions (Fig. 7 b). Although there was some overlap, cells in the densocellular region that projected to the VMS were generally located more medial and dorsal to those that projected to the DLS. Figure7 c illustrates the relationship between VMS efferent fibers and cells that project to the DLS. VMS efferent fibers terminated in the region of the densocellular group, which contained some cells that project to the DLS. This densocellular group was not within the efferent fibers originating in the DLS. Cases in which an anterograde tracer was injected into the ventral striatum and a retrograde tracer into the dorsolateral striatum show some labeled cells in the densocellular part of the ventral tier embedded in anterograde-labeled fibers from the ventral striatum (Fig. 7 c–e). In contrast, there was no overlap between the efferent projection from the DLS and the afferent projection to the VMS (Fig. 7 f). This is supported by the fact that after retrograde injections into the VMS, there were no labeled cells embedded in efferent fibers from the DLS.
Figure 8 illustrates the relationship of the CS pathways to the VMS and DLS circuits. The majority of labeled cells in the central densocellular region were derived from the CS injection sites. Fibers from the VMS that projected into the densocellular region overlapped extensively with the region of cells that projected to the CS (Fig. 8 a). This was particularly true of the dorsal population of CS-projecting cells that did not receive input from the CS. In contrast, there was little overlap between cells that projected to the VMS and efferent fibers of the CS (Fig. 8 b). However, efferent fibers from the CS that terminated ventral to the population of cells that projected reciprocally to the CS did overlap with cells that projected to the DLS (Fig. 8 c).
Distribution of labeled cells
After retrograde tracer injections throughout the midbrain, the densest distribution of labeled cells was found in the VMS. In contrast, the DLS had the fewest labeled cells (Fig.9 a). Injection sites throughout the dorsal and medial midbrain, including the dorsal tier and dorsal densocellular region, labeled cells in the VMS (Fig.9 a, open circles and light gray dots). The shell region was extensively labeled only after an injection site located at the midline, in the VTA (Fig. 9 b,stars). Although some labeled cells were found outside the shell region, they were mostly confined to the shell and extended only a short distance in the rostral striatum, with few labeled cells caudal to the anterior commissure. Ventromedial and dorsal injection sites labeled cells primarily in the core, outside of the shell, with some patches of labeled cells within the shell. Comparing a ventromedial injection site with one centered in the VTA, we found that labeled cells after the ventral sites were located primarily in the core outside the shell and showed a sharp contrast with the relatively unlabeled shell region (Fig. 9 b). All cases that labeled cells in the ventral striatum also labeled cells in the ventromedial part of the body and tail of the caudate nucleus. The ventral injection sites did not result in labeled VMS cells.
The CS contained labeled cells after injection sites that were located in the central densocellular region (Fig. 9 a,light and dark gray dots). Two ventral injection sites only labeled cells in the CS without tracer-positive cells in the VMS (Fig. 9 a, dark gray dots). Four centrally located sites had labeled cells primarily in the CS but extended into the VMS. There were labeled cells in the DLS after only four ventral injection sites. Combination of the four cases showed that the number and distribution of labeled striatal cells were relatively sparse in the DLS, with large areas lacking labeled cells (Fig.9 a, dark gray and black dots). Three injection sites included both the densocellular region and the cell columns and labeled both the DLS and the CS. Only one case had an injection site placed in the ventral pars reticulata. This injection site labeled cells only in the DLS (Fig. 9 a, black dots).
Location of fiber distributions in the striatum
Figure 9 c illustrates the distribution of the combined chartings of labeled fibers throughout the striatum after anterograde tracer injections into the midbrain. Eight injection sites were centered in different parts of the densocellular region: five included the dorsal densocellular region and parts of the dorsal tier, and three sites were placed more ventrally. Overall, fibers were distributed throughout the CS and the DLS. The VMS received the most limited projection.
Taken together, retrograde tracer sites centered in the densocellular region labeled cells primarily in the VMS and CS, with few labeled cells in the DLS. Conversely, anterograde injections centered in the densocellular region labeled fibers primarily in the CS and the DLS with relatively few fibers in the VMS. Injections of the bi-directional tracer LY into the densocellular region illustrates this projection pattern (Fig. 9 d). There were relatively few labeled fibers in the VMS. Patches of fibers were concentrated in the CS, with some also in the DLS (Fig. 9 e). In contrast, the distribution of labeled cells was most dense within the VMS (Fig.9 f). There were some clusters of cells in the CS and no labeled cells in the DLS.
Projections from the VMS terminate widely throughout the midbrain, and even the shell has an extensive medial/lateral terminal field (Fig.4 a,c). The retrograde experiments in the midbrain confirmed these findings. After most injection site locations, cells were labeled in the VMS. In contrast the DLS has a limited projection to the midbrain as evidenced by both the anterograde and retrograde experiments. The distribution of fibers and cells after anterograde and retrograde injections into the CS and midbrain, respectively, resulted in a pattern consistent with a projection to the CS primarily from the densocellular region and the dorsal part of the pars reticulata. Thus, with the exception of two retrograde injection sites in the midbrain, one placed into the ventral cell columns and one into the VTA, all cases showed some retrograde labeling in the CS (Fig. 9 a). The shell received input primarily from the VTA. Cells projecting to the core also originated from the dorsal tier and to some extent from the dorsal part of the densocellular group. These results were confirmed by the anterograde tracer injections into the midbrain. Despite a wide range of injection site locations, there were few labeled fibers in the shell. There were more fibers distributed in the core, but these were also limited to a few injection site locations that included the dorsal tier and medial midbrain. In contrast, the CS receives input primarily from the ventral tier. Labeled cells projecting to the CS were widely distributed throughout a range of the densocellular group. Anterograde tracer injections into the midbrain confirmed these results. There were clusters of labeled fibers throughout the CS after many injection site locations into the midbrain. The ventral tier also projected to the DLS, with the cell columns projecting almost exclusively to the DLS.
Each striatal injection site location was classified by its cortical input to the region and analyzed with respect to its relationship to limbic, associative, and motor cortices. On the basis of these results, we divided the striatum into the VMS, CS, and DLS. Previous studies have demonstrated the inverse dorsal–ventral topography of the striatonigral projection (Szabo, 1967, 1970, 1980;Selemon and Goldman-Rakic, 1990; Lynd-Balta and Haber, 1994c; Parent and Hazrati, 1994; Deniau et al., 1996) and of the nigrostriatal projection (Carpenter and Peter, 1971; Parent et al., 1983; Lynd-Balta and Haber, 1994a). When considered separately, each limb of the system creates a loose topographic organization demonstrating that the VTA and medial SN are associated with the limbic system, and the lateral and ventral SN are related to the associative and motor striatal regions. However, the VMS and DLS have contrasting relationships with the midbrain in that they differ in their relative proportional contribution to each limb of the SNS projection. Efferent projections from the VMS terminate throughout an extensive region of the dorsal midbrain, whereas the DLS projection is relatively limited to the ventrolateral SN. The CS pathway occupies an intermediate position between the VMS and DLS in the striatonigral pathway. Likewise, the ratio of nigrostriatal projections varies for different striatal regions. The VMS receives the most limited projection from the midbrain, whereas the DLS receives the largest. These differences in proportions significantly alter their relationship to the midbrain. The VMS influences a wide range of dopamine neurons but is itself influenced by a relatively limited group of dopamine cells. On the other hand, the DLS influences a limited midbrain region but is affected by a relatively large midbrain region.
Three SNS components in the SN
A major finding was that for each striatal region, the SNS projection system contained three SN components: a dorsal group of cells that does not lie within its reciprocal terminal field, a group of cells that does lie within its reciprocal terminal field, and a ventral component composed of the efferent terminals that do not contain a reciprocally connected group of labeled cells. Although the overlap between labeled cells and terminals at the light microscopic level does not demonstrate a direct synaptic connection, it is likely that this close relationship does indicate that the terminals in the region convey information relevant to the cells either directly or indirectly. Likewise, the lack of an overlap between terminal fields and labeled cells in the other two components is not necessarily evidence for a lack of connectivity, which might occur on distal dendrites. However, afferents nearest the cell bodies and proximal dendrites will be electrotonically closer to the soma and therefore likely to exert a much greater influence on spike activity than if they terminate distally (Spruston et al., 1994; Magee and Johnston, 1997). Thus, the three components of the SNS system are likely to represent different levels of interaction. Where cells and terminals overlap there is likely to be a more direct interface, which we refer to as a reciprocal connection. We refer to the dorsal and ventral components in which the cells and terminals do not overlap as nonreciprocal components. These three components for each SNS projection system occupy a different position within the midbrain. The VMS system lies dorsomedially, the DLS system lies ventrolaterally, and the CS system is positioned between the two (Fig.10).
The limbic–motor interface
The observation that the dorsal striatum is modulated by the ventral striatum was a proposed mechanism by which limbic circuitry affects motor outcome directly (Nauta and Domesick, 1978; Nauta et al., 1978; Somogyi et al., 1981; Haber and Fudge, 1997). However, the definition of the dorsal striatum was broad and referred to the entire dorsal striatal area, outside the nucleus accumbens. In this study we have examined the SNS circuit based on its frontal cortical input, including associative areas of dorsolateral prefrontal cortex. The organization of cortical inputs to the striatum imposes a functional gradient from limbic to associative to motor domains. A similar gradient is imposed on the midbrain by virtue of the SNS system. The three components of each SNS system overlap each other, with the interface of CS projections positioned between the VMS and DLS (Fig.11). Interactions between functional regions of the striatum via the midbrain will therefore be most robust between adjacent striatal regions. On the basis of the large striatal area that receives nonlimbic and nonmotor input, the direct interaction between the limbic and motor systems is limited. With this arrangement, a model of a “one step” limbic to motor interface through the midbrain is unlikely to be a major connection.
An ascending midbrain spiral mediates limbic input to motor outcome
Rather than a direct limbic–motor connection, we propose that through these three midbrain components, information from the limbic system reaches the motor system through a series of connections (Fig.12). The shell receives forebrain input primarily from areas most closely associated with the limbic system and projects to the dorsal tier. However, its efferent projection also terminates lateral and ventral to the dorsal tier, in the dorsal densocellular region. This area of terminal projection does not project back to the shell but rather to the core. Through this connection, cortical information that influences the dorsal tier through the shell also modulates the densocellular region that projects to the core. The idea that the shell influences the core via a series of connections or loops has been supported by previous work suggesting that the limbic system influences frontal cortex through striatopallidal pathways (Zahm and Brog, 1992). The study presented here provides another route by which informa- tion from the shell is directed to the core. Projections to and from the core also form a reciprocal loop with the midbrain. In addition, the core projects ventral to its reciprocal component, which interfaces with the CS but not the core. The CS is reciprocally connected to the densocellular region but also projects to the ventral pars reticulata and the cell columns. The cell columns project to the DLS, with a reciprocal connection back to the cell columns and ventral pars reticulata. The confined distribution of efferent DLS fibers limits the influence of the motor striatum to a relatively small region involving the cell columns and the pars reticulata. Taken together, the interface between different striatal regions via the midbrain DA cells is organized in an ascending spiral interconnecting different functional regions of the striatum (Fig. 12). Thus, rather than a direct limbic–motor interface, information flows through several circuits to reach the motor striatum.
Most of the ventral midbrain cells that project to the striatum are dopaminergic and play a key role in the acquisition of newly acquired behaviors. These cells discharge when presented with relevant stimuli for which a response is required. However, when the animal is overtrained, cells no longer respond, suggesting that it is not the movement but the relevance of the stimulus that is important (Schultz, 1992; Schultz et al., 1993; Wilson et al., 1995; Richardson and Gratton, 1996). Although afferent control of the dopamine neurons arises from a number of structures (Smith and Grace, 1992), the striatum is a major source. The striatum inhibits neurons in both the pars compacta and the pars reticulata. However, stimulation of the striatum can also lead to an increase in dopamine firing through inhibition of GABAergic interneurons (or pars reticulata cells) that terminate on dopaminergic cells and dendrites, resulting in disinhibition of pars compacta cells (Francois et al., 1979; Grace and Bunney, 1979, 1995; Johnson and North, 1992). This dual effect of both inhibition and disinhibition may be an important mechanism underlying the ascending SNS spiral control of information flow. Striatal response to dopamine is heterogeneous and determined by multiple variables, including complex interactions between several receptors (Di Chiara et al., 1994; Starr, 1995; Arbuthnott and Wickens, 1996; Wickens et al., 1996; Calabresi et al., 1997). Converging evidence indicates that tonic release of dopamine attenuates medium spiny neuronal response, whereas phasic release potentiates striatal response (Cepeda and Levine, 1998). In this way dopamine can both inhibit background corticostriatal input and facilitate (and therefore “focus”) specific corticostriatal synaptic transmission. If the reciprocal component of each limb of the spiral terminated directly on a dopamine cell, it would result in inhibition of dopamine burst firing (Fig. 12). Conversely, the nonreciprocal feedforward component of the ascending spiral might terminate on GABAergic interneurons and result in disinhibition and an increase of burst firing. Each component of information (from limbic to motor outcome) would send an inhibitory feedback response but facilitate transfer of information to the next step in the spiral (via disinhibition). Because information about potential reward of a specific behavior from the shell is conveyed to the midbrain, it would inhibit additional information flow from the shell via the reciprocal connection. The nonreciprocal feedforward projection, terminating in proximity to cells projecting to the core, would increase DA burst firing in the core via disinhibition. The reciprocal projection to the core also inhibits its midbrain feedback, but via the GABAergic interneuron it disinhibits cells projecting to the CS. Thus information transfer continues from the core to the CS, through the CS to the DLS and final motor outcome (Fig. 12).
The basal ganglia link between motivation and motor outcomes has focused primarily on pathways of the nucleus accumbens (Mogenson et al., 1980, 1993; Groenewegen et al., 1996). Behavioral studies of dopamine pathways have lead to the association of the mesolimbic pathway and nigrostriatal pathway with reward and motor activity, respectively. Although the role of dopamine and reward is well established (Wise and Rompre, 1989), its primary function is to direct attention to important stimuli likely to bring about a desired outcome (Ljungberg et al., 1992; Schultz et al., 1997). This requires processing a complex chain of events beginning with motivation and proceeding through cognitive processing that shapes final motor outcomes, a sequence reflected in the feedforward organization of the SNS connections.
This work was supported by National Institutes of Health Grants MH45573 and NS22511 (S.N.H.) and MH11661 (N.R.M.).
Correspondence should be addressed to Dr. Suzanne N. Haber, Department of Neurobiology and Anatomy, University of Rochester School of Medicine, 601 Elmwood Avenue, Rochester, NY 14642. E-mail:.